Posts in category: Amazon AWS

This post has been co-written with Artem Sysuev, Danny Portman, Matúš Chládek, and Saurabh Gupta from Zeta Global.

Zeta Global is a leading data-driven, cloud-based marketing technology company that empowers enterprises to acquire, grow and retain customers. The company’s Zeta Marketing Platform (ZMP) is the largest omnichannel marketing platform with identity data at its core. The ZMP analyzes billions of structured and unstructured data points to predict consumer intent by using sophisticated artificial intelligence (AI) to personalize experiences at scale. For more information, see Zeta Global’s home page.

What Zeta has accomplished in AI/ML

In the fast-evolving landscape of digital marketing, Zeta Global stands out with its groundbreaking advancements in artificial intelligence. Zeta’s AI innovations over the past few years span 30 pending and issued patents, primarily related to the application of deep learning and generative AI to marketing technology. Using AI, Zeta Global has revolutionized how brands connect with their audiences, offering solutions that aren’t just innovative, but also incredibly effective. As an early adopter of large language model (LLM) technology, Zeta released Email Subject Line Generation in 2021. This tool enables marketers to craft compelling email subject lines that significantly boost open rates and engagement, tailored perfectly to the audience’s preferences and behaviors.

Further expanding the capabilities of AI in marketing, Zeta Global has developed AI Lookalikes. This technology allows companies to identify and target new customers who closely resemble their best existing customers, thereby optimizing marketing efforts and improving their return on investment (ROI). The backbone of these advancements is ZOE, Zeta’s Optimization Engine. ZOE is a multi-agent LLM application that integrates with multiple data sources to provide a unified view of the customer, simplify analytics queries, and facilitate marketing campaign creation. Together, these AI-driven tools and technologies aren’t just reshaping how brands perform marketing tasks; they’re setting new benchmarks for what’s possible in customer engagement.

In addition to its groundbreaking AI innovations, Zeta Global has harnessed Amazon Elastic Container Service (Amazon ECS) with AWS Fargate to deploy a multitude of smaller models efficiently.

Zeta’s AI innovation is powered by a proprietary machine learning operations (MLOps) system, developed in-house.

Context

In early 2023, Zeta’s machine learning (ML) teams shifted from traditional vertical teams to a more dynamic horizontal structure, introducing the concept of pods comprising diverse skill sets. This paradigm shift aimed to accelerate project delivery by fostering collaboration and synergy among teams with varied expertise. The need for a centralized MLOps platform became apparent as ML and AI applications proliferated across various teams, leading to a maze of maintenance complexities and hindering knowledge transfer and innovation.

To address these challenges, the organization developed an MLOps platform based on four key open-source tools: Airflow, Feast, dbt, and MLflow. Hosted on Amazon ECS with tasks run on Fargate, this platform streamlines the end-to-end ML workflow, from data ingestion to model deployment. This blog post delves into the details of this MLOps platform, exploring how the integration of these tools facilitates a more efficient and scalable approach to managing ML projects.

Architecture overview

Our MLOps architecture is designed to automate and monitor all stages of the ML lifecycle. At its core, it integrates:

• Airflow for workflow orchestration
• Feast for feature management
• dbt for accelerated data transformation
• MLflow for experiment tracking and model management

These components interact within the Amazon ECS environment, providing a scalable and serverless platform where ML workflows are run in containers using Fargate. This setup not only simplifies infrastructure management, but also ensures that resources are used efficiently, scaling up or down as needed.

The following figure shows the MLOps architecture.

Architectural deep dive

The following details dive deep into each of the components used in this architecture.

Airflow for workflow orchestration

Airflow schedules and manages complex workflows, defining tasks and dependencies in Python code. An example direct acyclic graph (DAG) might automate data ingestion, processing, model training, and deployment tasks, ensuring that each step is run in the correct order and at the right time.

Though it’s worth mentioning that Airflow isn’t used at runtime as is usual for extract, transform, and load (ETL) tasks. Every Airflow task calls Amazon ECS tasks with some overrides. Additionally, we’re using a custom Airflow operator called ECSTaskLogOperator that allows us to process Amazon CloudWatch logs using downstream systems.

model_training = ECSTaskLogOperator(
task_id= <...>,
task_definition= <...>,
cluster= <...>,
launch_type="FARGATE",
aws_conn_id= <...>,
overrides={
"containerOverrides": [
{
"name": " <...> ",
"environment": [
{
"name": "MLFLOW_TRACKING_URI",
"value": "<...>"
},
],
"command": ["mlflow", "run", <...>]
}
],
}

Feast for feature management

Feast acts as a central repository for storing and serving features, ensuring that models in both training and production environments use consistent and up-to-date data. It simplifies feature access for model training and inference, significantly reducing the time and complexity involved in managing data pipelines.

Additionally, Feast promotes feature reuse, so the time spent on data preparation is reduced greatly.

from datetime import timedelta
from feast import Entity, FeatureView, FeatureService, Field, SnowflakeSource
from feast.types import Float64

entities = [
Entity(name="site_id", join_keys=["SITE_ID"]),
Entity(name="user_id", join_keys=["USER_ID"]),
]

def create_feature_view(name, table, field_name, schema_name):
return FeatureView(
name=name,
entities=entities,
ttl=timedelta(days=30),
schema=[Field(name=field_name, dtype=Float64)],
source=SnowflakeSource(
database="<...>", schema="<...>", table=table, 
timestamp_field="<...>"
),
tags="<...>",
)

feature_view_1 = create_feature_view("<...>")
feature_view_2 = create_feature_view("<...>")

my_feature_service = FeatureService(
name="my_feature_servic ",
features=[feature_view_1, feature_view_1],
description=""" 

This is my Feature Service 

""",
owner="<...>",
)

dbt for data transformation

dbt is used for transforming data within the data warehouse, allowing data teams to define complex data models in SQL. It promotes a disciplined approach to data modeling, making it easier to ensure data quality and consistency across the ML pipelines. Moreover, it provides a straightforward way to track data lineage, so we can foresee which datasets will be affected by newly introduced changes. The following figure shows schema definition and model which reference it.

MLflow for experiment tracking and model management

MLflow tracks experiments and manages models. It provides a unified interface for logging parameters, code versions, metrics, and artifacts, making it easier to compare experiments and manage the model lifecycle.

Similarly to Airflow, MLflow is also used just partially. The main parts we use are tracking the server and model registry. From our experience, artifact server has some limitations, such as limits on artifact size (because of sending it using REST API). As a result, we opted to use it only partially.

We don’t extensively use the deployment capabilities of MLflow, because in our current setup, we build custom inference containers.

Hosting on Amazon ECS with Fargate

Amazon ECS offers a highly scalable and secure environment for running containerized applications. Fargate eliminates the need for managing underlying infrastructure, allowing us to focus solely on deploying and running the containers. This abstraction layer simplifies the deployment process, enabling seamless scaling based on workload demands while optimizing resource utilization and cost efficiency.

We found it optimal to run on Fargate components of our ML workflows that don’t require GPUs or distributed processing. These include dbt pipelines, data gathering jobs, training, evaluation, and batch inference jobs for smaller models.

Furthermore, Amazon ECS and Fargate seamlessly integrate with other AWS services, such as Amazon Elastic Container Registry (Amazon ECR) for container image management and AWS Systems Manager Parameter Store for securely storing and managing secrets and configurations. Using Parameter Store, we can centralize configuration settings, such as database connection strings, API keys, and environment variables, eliminating the need for hardcoding sensitive information within container images. This enhances security and simplifies maintenance, because secrets and configuration values can be dynamically retrieved by containers at runtime, ensuring consistency across deployments.

Moreover, integrating Amazon ECS and Fargate with CloudWatch enables comprehensive monitoring and logging capabilities for containerized tasks. This can be achieved by enabling the awslogs log driver within the logConfiguration parameters of the task definitions.

Why ECS with Fargate is the solution of choice

1. Serverless model:

• • No infrastructure management: With Fargate, we don’t need to provision, configure, or manage servers. This simplifies operations and reduces the operational overhead, allowing teams to focus on developing and deploying applications.
  • Automatic scaling: Fargate automatically scales our applications based on demand, ensuring optimal performance without manual intervention.

2. Cost efficiency:

• • Pay-as-we-go: Fargate charges are based on the resources (vCPU and memory) that the containers use. This model can be more cost-effective compared to maintaining idle resources.
  • No over provisioning: Because we only pay for what we use, there’s no need to over-provision resources, which can lead to cost savings.

3. Enhanced security:

• • Isolation: Each Fargate task runs in its own isolated environment, improving security. There’s no sharing of underlying compute resources with other tenants.

4. Integration with the AWS ecosystem:

• • Seamless AWS service integration: Amazon ECS with Fargate integrates well with other AWS services such as Amazon Simple Storage Service (Amazon S3), Amazon Relational Database Service (Amazon RDS), and AWS Lambda, creating a cohesive ecosystem for deploying and managing applications.

Configuring Amazon ECS with Fargate for ML workloads

Configuring Amazon ECS with Fargate for ML workloads involves the following steps.

1. Docker images: ML models and applications are containerized using Docker. This includes all dependencies, libraries, and configurations needed to run the ML workload.
2. Creating task definitions:

• • Define resources: Create an Amazon ECS task definition specifying the Docker image, required vCPU, memory, and other configurations.
  • Environment variables: Set environment variables, such as model paths, API keys, and other necessary parameters.

3. IAM roles: Assign appropriate AWS Identity and Access Management (IAM) roles to the tasks for accessing other AWS resources securely.
4. Logging using CloudWatch: Use CloudWatch for logging and monitoring the performance and health of ML workloads.

Future development and addressing emerging challenges

As the field of MLOps continues to evolve, it’s essential to anticipate and address upcoming challenges to ensure that the platform remains efficient, scalable, and user-friendly. Two primary areas of future development for our platform include:

1. Enhancing bring your own model (BYOM) capabilities for external clients
2. Reducing the learning curve for data scientists

This section outlines those challenges and proposes directions for future enhancements.

Enhancing BYOM capabilities

As machine learning becomes more democratized, there is a growing need for platforms to easily integrate models developed externally by Zeta’s clients.

Future directions:

• Developing standardized APIs: Implement APIs that allow for easy integration of external models, regardless of the framework or language they were developed in. This would involve creating a set of standardized interfaces for model ingestion, validation, and deployment.
• Creating a model adapter framework: Design a framework that can adapt external models to be compatible with the platform’s infrastructure, ensuring that they can be managed, tracked, and deployed just like internally developed models.
• Enhancing documentation and support: Provide comprehensive documentation and support resources to guide external clients through the BYOM process, including best practices for model preparation, integration, and optimization.

Reducing the learning curve for data scientists

The incorporation of multiple specialized tools (Airflow, Feast, dbt, and MLflow) into the MLOps pipeline can present a steep learning curve for data scientists, potentially hindering their productivity and the overall efficiency of the ML development process.

Future directions:

We’ll do the following to help reduce the learning curve:

• Creating unified interfaces: Develop a unified interface, including UI, API, and SDK, that abstracts away the complexities of interacting with each tool individually. This interface could provide simplified workflows, automating routine tasks and presenting a cohesive view of the entire ML lifecycle.
• Offering comprehensive training and resources: Invest in training programs and resources tailored to data scientists at different skill levels. This could include interactive tutorials, workshops, and detailed case studies showcasing real-world applications of the platform.

Conclusion

Integrating Airflow, Feast, dbt, and MLflow into an MLOps platform hosted on Amazon ECS with AWS Fargate presents a robust solution for managing the ML lifecycle. This setup not only streamlines operations but also enhances scalability and efficiency, allowing data science teams to focus on innovation rather than infrastructure management.

Additional Resources

For those looking to dive deeper, we recommend exploring the official documentation and tutorials for each tool: Airflow, Feast, dbt, MLflow) and Amazon ECS. These resources are invaluable for understanding the capabilities and configurations of each component in our MLOps platform.

About the authors

Varad Ram holds the position of Senior Solutions Architect at Amazon Web Services. He possesses extensive experience encompassing application development, cloud migration strategies, and information technology team management. Recently, his primary focus has shifted towards assisting clients in navigating the process of productizing generative artificial intelligence use cases.

Artem Sysuev is a Lead Machine Learning Engineer at Zeta, passionate about creating efficient, scalable solutions. He believes that effective processes are key to success, which led him to focus on both machine learning and MLOps. Starting with machine learning, Artem developed skills in building predictive models. Over time, he saw the need for strong operational frameworks to deploy and maintain these models at scale, which drew him to MLOps. At Zeta, he drives innovation by automating workflows and improving collaboration, ensuring smooth integration of machine learning models into production systems.

Saurabh Gupta is a Principal Engineer at Zeta Global. He is passionate about machine learning engineering, distributed systems, and big-data technologies. He has built scalable platforms that empower data scientists and data engineers, focusing on low-latency, resilient systems that streamline workflows and drive innovation. He holds a B.Tech degree in Electronics and Communication Engineering from the Indian Institute of Technology (IIT), Guwahati, and has deep expertise in designing data-driven solutions that support advanced analytics and machine learning initiatives.

Matúš Chládek is a Senior Engineering Manager for ML Ops at Zeta Global. With a career that began in Data Science, Matúš has developed a strong foundation in analytics and machine learning. Over the years, Matúš transitioned into more engineering-focused roles, eventually becoming a Machine Learning Engineer before moving into Engineering Management. Matúš’s leadership focuses on building robust, scalable infrastructure that streamlines workflows and supports rapid iteration and production-ready delivery of machine learning projects. Matúš is passionate about driving innovation at the intersection of Data Science and Engineering, making advanced analytics accessible and scalable for internal users and clients alike.

Dr. Danny Portman is a recognized thought leader in AI and machine learning, with over 30 patents focused on Deep Learning and Generative AI applications in advertising and marketing technology. He holds a Ph.D. in Computational Physics, specializing in high-performance computing models for simulating complex astrophysical systems. With a strong background in quantitative research, Danny brings a wealth of experience in applying data-driven approaches to solve problems across various sectors. As VP of Data Science and Head of AI/ML at Zeta Global, Dr. Portman leads the development of AI-driven products and strategies, and spearheads the company’s cutting-edge Generative AI R&D efforts to deliver innovative solutions for marketers.

Read More

The post is co-written with Michael Shaul and Sasha Korman from NetApp.

Generative artificial intelligence (AI) applications are commonly built using a technique called Retrieval Augmented Generation (RAG) that provides foundation models (FMs) access to additional data they didn’t have during training. This data is used to enrich the generative AI prompt to deliver more context-specific and accurate responses without continuously retraining the FM, while also improving transparency and minimizing hallucinations.

In this post, we demonstrate a solution using Amazon FSx for NetApp ONTAP with Amazon Bedrock to provide a RAG experience for your generative AI applications on AWS by bringing company-specific, unstructured user file data to Amazon Bedrock in a straightforward, fast, and secure way.

Our solution uses an FSx for ONTAP file system as the source of unstructured data and continuously populates an Amazon OpenSearch Serverless vector database with the user’s existing files and folders and associated metadata. This enables a RAG scenario with Amazon Bedrock by enriching the generative AI prompt using Amazon Bedrock APIs with your company-specific data retrieved from the OpenSearch Serverless vector database.

When developing generative AI applications such as a Q&A chatbot using RAG, customers are also concerned about keeping their data secure and preventing end-users from querying information from unauthorized data sources. Our solution also uses FSx for ONTAP to allow users to extend their current data security and access mechanisms to augment model responses from Amazon Bedrock. We use FSx for ONTAP as the source of associated metadata, specifically the user’s security access control list (ACL) configurations attached to their files and folders and populate that metadata into OpenSearch Serverless. By combining access control operations with file events that notify the RAG application of new and changed data on the file system, our solution demonstrates how FSx for ONTAP enables Amazon Bedrock to only use embeddings from authorized files for the specific users that connect to our generative AI application.

AWS serverless services make it straightforward to focus on building generative AI applications by providing automatic scaling, built-in high availability, and a pay-for-use billing model. Event-driven compute with AWS Lambda is a good fit for compute-intensive, on-demand tasks such as document embedding and flexible large language model (LLM) orchestration, and Amazon API Gateway provides an API interface that allows for pluggable frontends and event-driven invocation of the LLMs. Our solution also demonstrates how to build a scalable, automated, API-driven serverless application layer on top of Amazon Bedrock and FSx for ONTAP using API Gateway and Lambda.

Solution overview

The solution provisions an FSx for ONTAP Multi-AZ file system with a storage virtual machine (SVM) joined to an AWS Managed Microsoft AD domain. An OpenSearch Serverless vector search collection provides a scalable and high-performance similarity search capability. We use an Amazon Elastic Compute Cloud (Amazon EC2) Windows server as an SMB/CIFS client to the FSx for ONTAP volume and configure data sharing and ACLs for the SMB shares in the volume. We use this data and ACLs to test permissions-based access to the embeddings in a RAG scenario with Amazon Bedrock.

The embeddings container component of our solution is deployed on an EC2 Linux server and mounted as an NFS client on the FSx for ONTAP volume. It periodically migrates existing files and folders along with their security ACL configurations to OpenSearch Serverless. It populates an index in the OpenSearch Serverless vector search collection with company-specific data (and associated metadata and ACLs) from the NFS share on the FSx for ONTAP file system.

The solution implements a RAG Retrieval Lambda function that allows RAG with Amazon Bedrock by enriching the generative AI prompt using Amazon Bedrock APIs with your company-specific data and associated metadata (including ACLs) retrieved from the OpenSearch Serverless index that was populated by the embeddings container component. The RAG Retrieval Lambda function stores conversation history for the user interaction in an Amazon DynamoDB table.

End-users interact with the solution by submitting a natural language prompt either through a chatbot application or directly through the API Gateway interface. The chatbot application container is built using Streamlit and fronted by an AWS Application Load Balancer (ALB). When a user submits a natural language prompt to the chatbot UI using the ALB, the chatbot container interacts with the API Gateway interface that then invokes the RAG Retrieval Lambda function to fetch the response for the user. The user can also directly submit prompt requests to API Gateway and obtain a response. We demonstrate permissions-based access to the RAG documents by explicitly retrieving the SID of a user and then using that SID in the chatbot or API Gateway request, where the RAG Retrieval Lambda function then matches the SID to the Windows ACLs configured for the document. As an additional authentication step in a production environment, you may want to also authenticate the user against an identity provider and then match the user against the permissions configured for the documents.

The following diagram illustrates the end-to-end flow for our solution. We start by configuring data sharing and ACLs with FSx for ONTAP, and then these are periodically scanned by the embeddings container. The embeddings container splits the documents into chunks and uses the Amazon Titan Embeddings model to create vector embeddings from these chunks. It then stores these vector embeddings with associated metadata in our vector database by populating an index in a vector collection in OpenSearch Serverless. The following diagram illustrates the end-to-end flow.

The following architecture diagram illustrates the various components of our solution.

Prerequisites

Complete the following prerequisite steps:

1. Make sure you have model access in Amazon Bedrock. In this solution, we use Anthropic Claude v3 Sonnet on Amazon Bedrock.
2. Install the AWS Command Line Interface (AWS CLI).
3. Install Docker.
4. Install Terraform.

Deploy the solution

The solution is available for download on this GitHub repo. Cloning the repository and using the Terraform template will provision all the components with their required configurations.

1. Clone the repository for this solution:

   sudo yum install -y unzip
   git clone https://github.com/aws-samples/genai-bedrock-fsxontap.git
   cd genai-bedrock-fsxontap/terraform

2. From the terraform folder, deploy the entire solution using Terraform:

   terraform init
   terraform apply -auto-approve

This process can take 15–20 minutes to complete. When finished, the output of the terraform commands should look like the following:

api-invoke-url = "https://9ng1jjn8qi.execute-api.<region>.amazonaws.com/prod"
fsx-management-ip = toset([
"198.19.255.230",])
fsx-secret-id = "arn:aws:secretsmanager:<region>:<account-id>:secret:AmazonBedrock-FSx-NetAPP-ONTAP-a2fZEdIt-0fBcS9"
fsx-svm-smb-dns-name = "BRSVM.BEDROCK-01.COM"
lb-dns-name = "chat-load-balancer-2040177936.<region>.elb.amazonaws.com"

Load data and set permissions

To test the solution, we will use the EC2 Windows server (ad_host) mounted as an SMB/CIFS client to the FSx for ONTAP volume to share sample data and set user permissions that will then be used to populate the OpenSearch Serverless index by the solution’s embedding container component. Perform the following steps to mount your FSx for ONTAP SVM data volume as a network drive, upload data to this shared network drive, and set permissions based on Windows ACLs:

1. Obtain the ad_host instance DNS from the output of your Terraform template.
2. Navigate to AWS Systems Manager Fleet Manager on your AWS console, locate the ad_host instance and follow instructions here to login with Remote Desktop. Use the domain admin user bedrock-01Admin and obtain the password from AWS Secrets Manager. You can find the password using the Secrets Manager fsx-secret-id secret id from the output of your Terraform template.
3. To mount an FSx for ONTAP data volume as a network drive, under This PC, choose (right-click) Network and then choose Map Network drive.
4. Choose the drive letter and use the FSx for ONTAP share path for the mount
   (\<svm>.<domain >c$<volume-name>):
   
5. Upload the Amazon Bedrock User Guide to the shared network drive and set permissions to the admin user only (make sure that you disable inheritance under Advanced):
6. Upload the Amazon FSx for ONTAP User Guide to the shared drive and make sure permissions are set to Everyone:
7. On the ad_host server, open the command prompt and enter the following command to obtain the SID for the admin user:

   wmic useraccount where name='Admin' get sid

Test permissions using the chatbot

To test permissions using the chatbot, obtain the lb-dns-name URL from the output of your Terraform template and access it through your web browser:

For the prompt query, ask any general question on the FSx for ONTAP user guide that is available for access to everyone. In our scenario, we asked “How can I create an FSx for ONTAP file system,” and the model replied back with detailed steps and source attribution in the chat window to create an FSx for ONTAP file system using the AWS Management Console, AWS CLI, or FSx API:

Now, let’s ask a question about the Amazon Bedrock user guide that is available for admin access only. In our scenario, we asked “How do I use foundation models with Amazon Bedrock,” and the model replied with the response that it doesn’t have enough information to provide a detailed answer to the question.:

Use the admin SID on the user (SID) filter search in the chat UI and ask the same question in the prompt. This time, the model should reply with steps detailing how to use FMs with Amazon Bedrock and provide the source attribution used by the model for the response:

Test permissions using API Gateway

You can also query the model directly using API Gateway. Obtain the api-invoke-url parameter from the output of your Terraform template.

curl -v '<api-invoke-url>/bedrock_rag_retreival' -X POST -H 'content-type: application/json' -d '{"session_id": "1","prompt": "What is an FSxN ONTAP filesystem?", "bedrock_model_id": "anthropic.claude-3-sonnet-20240229-v1:0", "model_kwargs": {"temperature": 1.0, "top_p": 1.0, "top_k": 500}, "metadata": "NA", "memory_window": 10}'

Then invoke the API gateway with Everyone access for a query related to the FSx for ONTAP user guide by setting the value of the metadata parameter to NA to indicate Everyone access:

curl -v '<api-invoke-url>/bedrock_rag_retreival' -X POST -H 'content-type: application/json' -d '{"session_id": "1","prompt": "what is bedrock?", "bedrock_model_id": "anthropic.claude-3-sonnet-20240229-v1:0", "model_kwargs": {"temperature": 1.0, "top_p": 1.0, "top_k": 500}, "metadata": "S-1-5-21-4037439088-1296877785-2872080499-1112", "memory_window": 10}'

Cleanup

To avoid recurring charges, clean up your account after trying the solution. From the terraform folder, delete the Terraform template for the solution:

terraform apply --destroy

Conclusion

In this post, we demonstrated a solution that uses FSx for ONTAP with Amazon Bedrock and uses FSx for ONTAP support for file ownership and ACLs to provide permissions-based access in a RAG scenario for generative AI applications. Our solution enables you to build generative AI applications with Amazon Bedrock where you can enrich the generative AI prompt in Amazon Bedrock with your company-specific, unstructured user file data from an FSx for ONTAP file system. This solution enables you to deliver more relevant, context-specific, and accurate responses while also making sure only authorized users have access to that data. Finally, the solution demonstrates the use of AWS serverless services with FSx for ONTAP and Amazon Bedrock that enable automatic scaling, event-driven compute, and API interfaces for your generative AI applications on AWS.

For more information about how to get started building with Amazon Bedrock and FSx for ONTAP, refer to the following resources:

• Amazon Bedrock Workshop GitHub repo
• Amazon FSx for NetApp ONTAP File Storage Workshop GitHub repo
• NetApp helps customers unlock the full potential of GenAI with BlueXP workload factory and Amazon Bedrock

About the authors

Kanishk Mahajan is Principal, Solutions Architecture at AWS. He leads cloud transformation and solution architecture for ISV customers and partner at AWS. Kanishk specializes in containers, cloud operations, migrations and modernizations, AI/ML, resilience and security and compliance. He is a Technical Field Community (TFC) member in each of those domains at AWS.

Michael Shaul is a Principal Architect at NetApp’s office of the CTO. He has over 20 years of experience building data management systems, applications, and infrastructure solutions. He has a unique in-depth perspective on cloud technologies, builder, and AI solutions.

Sasha Korman is a tech visionary leader of dynamic development and QA teams across Israel and India. With 14-years at NetApp that began as a programmer, his hands-on experience and leadership have been pivotal in steering complex projects to success, with a focus on innovation, scalability, and reliability.

Read More

AWS DeepComposer was first introduced during AWS re:Invent 2019 as a fun way for developers to compose music by using generative AI. AWS DeepComposer was the world’s first machine learning (ML)-enabled keyboard for developers to get hands-on—literally—with a musical keyboard and the latest ML techniques to compose their own music.

After careful consideration, we have made the decision to end support for AWS DeepComposer, effective September 17, 2025. With your help and feedback, our portfolio of products and services has grown to include new tools for developers to get hands-on with AI and ML. Amazon PartyRock, for example, is a generative AI playground for intuitive, code-free help in building web applications.

If you have data stored on the AWS DeepComposer console, you will be able to use AWS DeepComposer as normal until September 17, 2025, when support for the service will end. After this date, you will no longer be able to use AWS DeepComposer through the AWS Management Console, manage AWS DeepComposer devices, or access any compositions or models you have created. Until then, you can continue to work on your compositions or models and export those you would like to keep by using the step-by-step guide in the AWS DeepComposer FAQs.

If you have additional questions, please read our FAQs or contact us.

About the author

Kanchan Jagannathan is a Sr. Program Manager in the AWS AI Devices team where he helps launches AWS devices into sales channel and also oversees the Service Availability Change process for the team. He was a Program Manager for FC automation deployment and launches before joining AWS. Outside of work, he has begun to bravely endeavour camping with his 5-yr old and 1-yr old kids and enjoying the moments he gets to be with them.

Read More

Amazon Lookout for Equipment, the AWS machine learning (ML) service designed for industrial equipment predictive maintenance, will no longer be open to new customers effective October 17, 2024. Existing customers will be able to use the service (both using the AWS Management Console and API) as normal and AWS will continue to invest in security, availability, and performance improvements for Lookout for Equipment, but we do not plan to introduce new features for this service.

This post discusses how you can maintain access to Lookout for Equipment after it is closed to new customers and some alternatives to Lookout for Equipment.

Maintaining access to Lookout for Equipment

You’re considered an existing customer if you use the service, either through cloud training or cloud inferencing, any time in the 30 days prior to October 17, 2024 (September 17, 2024, through October 16, 2024). To maintain access to the service after October 17, 2024, you should complete one of the following tasks from the account for which you intend to maintain access:

• On the Lookout for Equipment console, start a new project and successfully complete a model training
• On the Lookout for Equipment console, open an existing project, schedule an inference for a given model, and run at least one inference
• Use Lookout for Equipment API calls CreateInferenceScheduler and StartInferenceScheduler (and StopInferenceScheduler when done)

For any questions or support needed, contact your assigned AWS Account Manager or Solutions Architect, or create a case from the AWS console.

Alternatives to Lookout for Equipment

If you’re interested in an alternative to Lookout for Equipment, AWS has options for both buyers and builders.

For an out-of-the-box solution, the AWS Partner Network offers solutions from multiple partners. You can browse solutions on the Asset Maintenance and Reliability page in the AWS Solutions Library. This approach provides a solution that addresses your use case without requiring you to have expertise in predictive maintenance, and typically provides the fastest time to value by using the specialized expertise of the AWS Partners.

If you prefer to build your own solution, AWS offers AI/ML tools and services to help you develop an AI-based predictive maintenance solution. Amazon SageMaker provides a set of tools to enable you to build, train, infer, and deploy ML models for your use case with fully managed infrastructure, tools, and workflows.

Summary

Although new customers will no longer have access to Lookout for Equipment after October 17, 2024, AWS offers a powerful set of AI/ML services and solutions in the form of SageMaker tools to build customer models, and also offers a range of solutions from partners through the AWS Partner Network. You should explore these options to determine what works best for your specific needs.

For more details, refer to the following resources:

• Amazon Lookout for Equipment Developer Guide
• Amazon SageMaker Developer Guide
• AWS Solutions Library

About the author

Stuart Gillen is a Sr. Product Manager, Lookout for Equipment, at AWS. Stuart has held a variety of roles in engineering management, business development, product management, and consulting. Most of his career has been focused on industrial applications specifically in reliability practices, maintenance systems, and manufacturing.
Stuart is the Product Manager for Lookout for Equipment at AWS where he utilizes his industrial and AI background in applications focusing on Predictive Maintenance and Condition Monitoring.

Read More

The clustered regularly interspaced short palindromic repeat (CRISPR) technology holds the promise to revolutionize gene editing technologies, which is transformative to the way we understand and treat diseases. This technique is based in a natural mechanism found in bacteria that allows a protein coupled to a single guide RNA (gRNA) strand to locate and make cuts in specific sites in the targeted genome. Being able to computationally predict the efficiency and specificity of gRNA is central to the success of gene editing.

Transcribed from DNA sequences, RNA is an important type of biological sequence of ribonucleotides (A, U, G, C), which folds into 3D structure. Benefiting from recent advance in large language models (LLMs), a variety of computational biology tasks can be solved by fine-tuning biological LLMs pre-trained on billions of known biological sequences. The downstream tasks on RNAs are relatively understudied.

In this post, we adopt a pre-trained genomic LLMs for gRNA efficiency prediction. The idea is to treat a computer designed gRNA as a sentence, and fine-tune the LLM to perform sentence-level regression tasks analogous to sentiment analysis. We used Parameter-Efficient Fine-Tuning methods to reduce the number of parameters and GPU usage for this task.

Solution overview

Large language models (LLMs) have gained a lot of interest for their ability to encode syntax and semantics of natural languages. The neural architecture behind LLMs are transformers, which are comprised of attention-based encoder-decoder blocks that generate an internal representation of the data they are trained from (encoder) and are able to generate sequences in the same latent space that resemble the original data (decoder). Due to their success in natural language, recent works have explored the use of LLMs for molecular biology information, which is sequential in nature.

DNABERT is a pre-trained transformer model with non-overlapping human DNA sequence data. The backbone is a BERT architecture made up of 12 encoding layers. The authors of this model report that DNABERT is able to capture a good feature representation of the human genome that enables state-of-the-art performance on downstream tasks like promoter prediction and splice/binding site identification. We decided to use this model as the foundation for our experiments.

Despite the success and popular adoption of LLMs, fine-tuning these models can be difficult because of the number of parameters and computation necessary for it. For this reason, Parameter-Efficient Fine-Tuning (PEFT) methods have been developed. In this post, we use one of these methods, called LoRA (Low-Rank Adaptation). We introduce the method in the following sections.

The following diagram is a representation of the Cas9 DNA target mechanism. The gRNA is the component that helps target the cleavage site.

The goal of this solution is to fine-tune a base DNABERT model to predict activity efficiency from different gRNA candidates. As such, our solution first takes gRNA data and processes it, as described later in this post. Then we use an Amazon SageMaker notebook and the Hugging Face PEFT library to fine-tune the DNABERT model with the processed RNA data. The label we want to predict is the efficiency score as it was calculated in experimental conditions testing with the actual RNA sequences in cell cultures. Those scores describe a balance between being able to edit the genome and not damage DNA that wasn’t targeted.

The following diagram illustrates the workflow of the proposed solution.

Prerequisites

For this solution, you need access to the following:

• A SageMaker notebook instance (we trained the model on an ml.g4dn.8xlarge instance with a single NVIDIA T4 GPU)
• transformers-4.34.1
• peft-0.5.0
• DNABERT 6

Dataset

For this post, we use the gRNA data released by researchers in a paper about gRNA prediction using deep learning. This dataset contains efficiency scores calculated for different gRNAs. In this section, we describe the process we followed to create the training and evaluation datasets for this task.

To train the model, you need a 30-mer gRNA sequence and efficiency score. A k-mer is a contiguous sequence of k nucleotide bases extracted from a longer DNA or RNA sequence. For example, if you have the DNA sequence “ATCGATCG” and you choose k = 3, then the k-mers within this sequence would be “ATC,” “TCG,” “CGA,” “GAT,” and “ATC.”

Efficiency score

Start with excel file 41467_2021_23576_MOESM4_ESM.xlsx from the CRISPRon paper in the Supplementary Data 1 section. In this file, the authors released the gRNA (20-mer) sequences and corresponding total_indel_eff scores. We specifically used the data from the sheet named spCas9_eff_D10+dox. We use the total_indel_eff column as the efficiency score.

Training and validation data

Given the 20-mers and the crispron scores (same as the total_indel_eff scores) from earlier, complete the following steps to put together the training and validation data:

1. Convert the sequences in the sheet “TRAP12K microarray oligos” into an .fa (fasta) file.
2. Run the script get_30mers_from_fa.py (from the CRISPRon GitHub repository) to obtain all possible 23-mers and 30-mers from the sequences obtained from Step 1.
3. Use the CRISPRspec_CRISPRoff_pipeline.py script (from the CRISPRon GitHub repository) to obtain the binding energy for the 23-mers obtained from Step 2. For more details on how to run this script, check out the code released by the authors of the CRISPRon paper(check the script CRISPRon.sh).
4. At this point, we have 23-mers along with the corresponding binding energy scores, and 20-mers along with the corresponding CRISPRon scores. Additionally, we have the 30-mers from Step 2.
5. Use the script prepare_train_dev_data.py (from our released code) to create training and validation splits. Running this script will create two files: train.csv and dev.csv.

The data looks something like the following:

id,rna,crisproff_score,crispron_score
seq2875_p_129,GTCCAGCCACCGAGACCCTGTGTATGGCAC,24.74484099890205,85.96491228
seq2972_p_129,AAAGGCGAAGCAGTATGTTCTAAAAGGAGG,17.216228493196073,94.81132075
. . .
. . .

Model architecture for gRNA encoding

To encode the gRNA sequence, we used the DNABERT encoder. DNABERT was pre-trained on human genomic data, so it’s a good model to encode gRNA sequences. DNABERT tokenizes the nucleotide sequence into overlapping k-mers, and each k-mer serves as a word in the DNABERT model’s vocabulary. The gRNA sequence is broken into a sequence of k-mers, and then each k-mer is replaced by an embedding for the k-mer at the input layer. Otherwise, the architecture of DNABERT is similar to that of BERT. After we encode the gRNA, we use the representation of the [CLS] token as the final encoding of the gRNA sequence. To predict the efficiency score, we use an additional regression layer. The MSE loss will be the training objective. The following is a code snippet of the DNABertForSequenceClassification model:

class DNABertForSequenceClassification(BertPreTrainedModel):
    def __init__(self, config):
        super().__init__(config)
        self.num_labels = config.num_labels
        self.config = config
        
        self.bert = BertModel(config)
        classifier_dropout = (
            config.classifier_dropout
            if config.classifier_dropout is not None
            else config.hidden_dropout_prob
        )
        self.dropout = nn.Dropout(classifier_dropout)
        self.classifier = nn.Linear(config.hidden_size, config.num_labels)
        # Initialize weights and apply final processing
        self.post_init()

    def forward(
        self,
        input_ids: Optional[torch.Tensor] = None,
        attention_mask: Optional[torch.Tensor] = None,
        token_type_ids: Optional[torch.Tensor] = None,
        position_ids: Optional[torch.Tensor] = None,
        head_mask: Optional[torch.Tensor] = None,
        inputs_embeds: Optional[torch.Tensor] = None,
        labels: Optional[torch.Tensor] = None,
        output_attentions: Optional[bool] = None,
        output_hidden_states: Optional[bool] = None,
        return_dict: Optional[bool] = None,
    ) -> Union[Tuple[torch.Tensor], SequenceClassifierOutput]:
        r"""
        labels (`torch.LongTensor` of shape `(batch_size,)`, *optional*):
            Labels for computing the sequence classification/regression loss. Indices should be in `[0, ...,
            config.num_labels - 1]`. If `config.num_labels == 1` a regression loss is computed (Mean-Square loss), If
            `config.num_labels > 1` a classification loss is computed (Cross-Entropy).
        """
        return_dict = (
            return_dict if return_dict is not None else self.config.use_return_dict
        )

        outputs = self.bert(
            input_ids,
            attention_mask=attention_mask,
            token_type_ids=token_type_ids,
            position_ids=position_ids,
            head_mask=head_mask,
            inputs_embeds=inputs_embeds,
            output_attentions=output_attentions,
            output_hidden_states=output_hidden_states,
            return_dict=return_dict,
        )
        print('bert outputs', outputs)
        pooled_output = outputs[1]
        pooled_output = self.dropout(pooled_output)
        logits = self.classifier(pooled_output)

        loss = None
        if labels is not None:
            if self.config.problem_type is None:
                if self.num_labels == 1:
                    self.config.problem_type = "regression"
                elif self.num_labels > 1 and (
                    labels.dtype == torch.long or labels.dtype == torch.int
                ):
                    self.config.problem_type = "single_label_classification"
                else:
                    self.config.problem_type = "multi_label_classification"

            if self.config.problem_type == "regression":
                loss_fct = MSELoss()
                if self.num_labels == 1:
                    loss = loss_fct(logits.squeeze(), labels.squeeze())
                else:
                    loss = loss_fct(logits, labels)
            elif self.config.problem_type == "single_label_classification":
                loss_fct = CrossEntropyLoss()
                loss = loss_fct(logits.view(-1, self.num_labels), labels.view(-1))
            elif self.config.problem_type == "multi_label_classification":
                loss_fct = BCEWithLogitsLoss()
                loss = loss_fct(logits, labels)
        if not return_dict:
            output = (logits,) + outputs[2:]
            return ((loss,) + output) if loss is not None else output

        return SequenceClassifierOutput(
            loss=loss,
            logits=logits,
            hidden_states=outputs.hidden_states,
            attentions=outputs.attentions,
        )

Fine-tuning and prompting genomic LLMs

Fine-tuning all the parameters of a model is expensive because the pre-trained model becomes much larger. LoRA is an innovative technique developed to address the challenge of fine-tuning extremely large language models. LoRA offers a solution by suggesting that the pre-trained model’s weights remain fixed while introducing trainable layers (referred to as rank-decomposition matrices) within each transformer block. This approach significantly reduces the number of parameters that need to be trained and lowers the GPU memory requirements, because most model weights don’t require gradient computations.

Therefore, we adopted LoRA as a PEFT method on the DNABERT model. LoRA is implemented in the Hugging Face PEFT library. When using PEFT to train a model with LoRA, the hyperparameters of the low rank adaptation process and the way to wrap base transformers models can be defined as follows:

from peft import LoraConfig

tokenizer = AutoTokenizer.from_pretrained(
        data_training_args.model_path,
        do_lower_case=False
    )
# DNABertForSequenceClassification is a model class for sequence classification task, which is built on top of the DNABert architecture.    
model = DNABertForSequenceClassification.from_pretrained(
        data_training_args.model_path,
        config=config
    )
    
# Define LoRA Config
LORA_R = 16
LORA_ALPHA = 16
LORA_DROPOUT = 0.05
peft_config = LoraConfig(
                     r=LORA_R, # the dimension of the low-rank matrices
                     lora_alpha=LORA_ALPHA, #scaling factor for the weight matrices
                     lora_dropout=LORA_DROPOUT, #dropout probability of the LoRA layers
                     bias="none",
                     task_type = 'SEQ_CLS'
    )
model = get_peft_model(model, peft_config)

Hold-out evaluation performances

We use RMSE, MSE, and MAE as evaluation metrics, and we tested with rank 8 and 16. Furthermore, we implemented a simple fine-tuning method, which is simply adding several dense layers after the DNABERT embeddings. The following table summarizes the results.

When rank=8, we have 296,450 trainable parameters, which is about 33% trainable of the whole. The performance metrics are “rmse”: 11.933, “mse”: 142.397, “mae”: 7.014.

When rank=16, we have 591,362 trainable parameters, which is about 66% trainable of the whole. The performance metrics are “rmse”: 13.039, “mse”: 170.010, “mae”: 7.157. There might have some overfitting issue here under this setting.

We also compare what happens when adding a few dense layers:

• After adding one dense layer, we have “rmse”: 15.435, “mse”: 238.265, “mae”: 9.351
• After adding three dense layers, we have “rmse”: 15.435, “mse”: 238.241, “mae”: 9.505

Lastly, we compare with the existing CRISPRon method. CRISPRon is a CNN based deep learning model. The performance metrics are “rmse”: 11.788, “mse”: 138.971, “mae”: 7.134.

As expected, LoRA is doing much better than simply adding a few dense layers. Although the performance of LoRA is a bit worse than CRISPRon, with thorough hyperparameter search, it is likely to outperform CRISPRon.

When using SageMaker notebooks, you have the flexibility to save the work and data produced during the training, turn off the instance, and turn it back on when you’re ready to continue the work, without losing any artifacts. Turning off the instance will keep you from incurring costs on compute you’re not using. We highly recommend only turning it on when you’re actively using it.

Conclusion

In this post, we showed how to use PEFT methods for fine-tuning DNA language models using SageMaker. We focused on predicting efficiency of CRISPR-Cas9 RNA sequences for their impact in current gene-editing technologies. We also provided code that can help you jumpstart your biology applications in AWS.

To learn more about the healthcare and life science space, refer to Run AlphaFold v2.0 on Amazon EC2 or fine-tuning Fine-tune and deploy the ProtBERT model for protein classification using Amazon SageMaker.

About the Authors

Siddharth Varia is an applied scientist in AWS Bedrock. He is broadly interested in natural language processing and has contributed to AWS products such as Amazon Comprehend. Outside of work, he enjoys exploring new places and reading. He got interested in this project after reading the book The Code Breaker.

Yudi Zhang is an Applied Scientist at AWS marketing. Her research interests are in the area of graph neural networks, natural language processing, and statistics.

Erika Pelaez Coyotl is a Sr Applied Scientist in Amazon Bedrock, where she’s currently helping develop the Amazon Titan large language model. Her background is in biomedical science, and she has helped several customers develop ML models in this vertical.

Zichen Wang is a Sr Applied Scientist in AWS AI Research & Education. He is interested in researching graph neural networks and applying AI to accelerate scientific discovery, specifically on molecules and simulations.

Rishita Anubhai is a Principal Applied Scientist in Amazon Bedrock. She has deep expertise in natural language processing and has contributed to AWS projects like Amazon Comprehend, Machine Learning Solutions Lab, and development of Amazon Titan models. She’s keenly interested in using machine learning research, specifically deep learning, to create tangible impact.

Read More

This post is co-written with Pradeep Prabhakaran from Cohere.

Retrieval Augmented Generation (RAG) is a powerful technique that can help enterprises develop generative artificial intelligence (AI) apps that integrate real-time data and enable rich, interactive conversations using proprietary data.

RAG allows these AI applications to tap into external, reliable sources of domain-specific knowledge, enriching the context for the language model as it answers user queries. However, the reliability and accuracy of the responses hinges on finding the right source materials. Therefore, honing the search process in RAG is crucial to boosting the trustworthiness of the generated responses.

RAG systems are important tools for building search and retrieval systems, but they often fall short of expectations due to suboptimal retrieval steps. This can be enhanced using a rerank step to improve search quality.

RAG is an approach that combines information retrieval techniques with natural language processing (NLP) to enhance the performance of text generation or language modeling tasks. This method involves retrieving relevant information from a large corpus of text data and using it to augment the generation process. The key idea is to incorporate external knowledge or context into the model to improve the accuracy, diversity, and relevance of the generated responses.

Workflow of RAG Orchestration

The RAG orchestration generally consists of two steps:

1. Retrieval – RAG fetches relevant documents from an external data source using the generated search queries. When presented with the search queries, the RAG-based application searches the data source for relevant documents or passages.
2. Grounded generation – Using the retrieved documents or passages, the generation model creates educated answers with inline citations using the fetched documents.

The following diagram shows the RAG workflow.

Document retrieval in RAG orchestration

One technique for retrieving documents in a RAG orchestration is dense retrieval, which is an approach to information retrieval that aims to understand the semantic meaning and intent behind user queries. Dense retrieval finds the closest documents to a user query in the embedding, as shown in the following screenshot.

The goal of dense retrieval is to map both the user queries and documents (or passages) into a dense vector space. In this space, the similarity between the query and document vectors can be computed using standard distance metrics like cosine similarity or euclidean distance. The documents that match closest to the semantic meaning of the user query based on the calculated distance metrics are then presented back to the user.

The quality of the final responses to search queries is significantly influenced by the relevance of the retrieved documents. While dense retrieval models are very efficient and can scale to large datasets, they struggle with more complex data and questions due to the simplicity of the method. Document vectors contain the meaning of text in a compressed representation—typically 786-1536 dimension vectors. This often results in loss of information because information is compressed into a single vector. When documents are retrieved during a vector search the most relevant information is not always presented at the top of the retrieval.

Boost search accuracy with Cohere Rerank

To address the challenges with accuracy, search engineers have used two-stage retrieval as a means of increasing search quality. In these two-stage systems, a first-stage model (an embedding model or retriever) retrieves a set of candidate documents from a larger dataset. Then, a second-stage model (the reranker) is used to rerank those documents retrieved by the first-stage model.

A reranking model, such as Cohere Rerank, is a type of model that will output a similarity score when given a query and document pair. This score can be used to reorder the documents that are most relevant to the search query. Among the reranking methodologies, the Cohere Rerank model stands out for its ability to significantly enhance search accuracy. The model diverges from traditional embedding models by employing deep learning to evaluate the alignment between each document and the query directly. Cohere Rerank outputs a relevance score by processing the query and document in tandem, which results in a more nuanced document selection process.

In the following example, the application was presented with a query: “When was the transformer paper coauthored by Aidan Gomez published?” The top-k with k = 6 returned the results shown in the image, in which the retrieved result set did contain the most accurate result, although it was at the bottom of the list. With k = 3, the most relevant document would not be included in the retrieved results.

Cohere Rerank aims to reassess and reorder the relevance of the retrieved documents based on additional criteria, such as semantic content, user intent, and contextual relevance, to output a similarity score. This score is then used to reorder the documents by relevance of the query. The following image shows reorder results using Rerank.

By applying Cohere Rerank after the first-stage retrieval, the RAG orchestration can gain the benefits of both approaches. While first-stage retrieval helps to capture relevant items based on proximity matches within the vector space, reranking helps optimize search according to results by guaranteeing contextually relevant results are surfaced to the top. The following diagram demonstrates this improved efficiency.

The latest version of Cohere Rerank, Rerank 3, is purpose-built to enhance enterprise search and RAG systems. Rerank 3 offers state-of-the-art capabilities for enterprise search, including:

• 4k context length to significantly improve search quality for longer documents
• Ability to search over multi-aspect and semi-structured data (such as emails, invoices, JSON documents, code, and tables)
• Multilingual coverage of more than 100 languages
• Improved latency and lower total cost of ownership (TCO)

The endpoint takes in a query and a list of documents, and it produces an ordered array with each document assigned a relevance score. This provides a powerful semantic boost to the search quality of any keyword or vector search system without requiring any overhaul or replacement.

Developers and businesses can access Rerank on Cohere’s hosted API and on Amazon SageMaker. This post offers a step-by-step walkthrough of consuming Cohere Rerank on Amazon SageMaker.

Solution overview

This solution follows these high-level steps:

1. Subscribe to the model package
2. Create an endpoint and perform real-time inference

Prerequisites

For this walkthrough, you must have the following prerequisites:

1. The cohere-aws notebook.

This is a reference notebook, and it cannot run unless you make changes suggested in the notebook. It contains elements that render correctly in the Jupyter interface, so you need to open it from an Amazon SageMaker notebook instance or in Amazon SageMaker Studio.

2. An AWS Identity and Access Management (IAM) role with the AmazonSageMakerFullAccess policy attached. To deploy this machine learning (ML) model successfully, choose one of the following options:
   1. If your AWS account does not have a subscription to Cohere Rerank 3 Model – Multilingual, your IAM role needs to have the following three permissions, and you need to have the authority to make AWS Marketplace subscriptions in the AWS account used:
      • aws-marketplace:ViewSubscriptions
      • aws-marketplace:Unsubscribe
      • aws-marketplace:Subscribe
   2. If your AWS account has a subscription to Cohere Rerank 3 Model – Multilingual, you can skip the instructions for subscribing to the model package.

Refrain from using full access in production environments. Security best practice is to opt for the principle of least privilege.

Implement Rerank 3 on Amazon SageMaker

To improve RAG performance using Cohere Rerank, use the instructions in the following sections.

Subscribe to the model package

To subscribe to the model package, follow these steps:

1. In AWS Marketplace, open the model package listing page Cohere Rerank 3 Model – Multilingual
2. Choose Continue to Subscribe.
3. On the Subscribe to this software page, review the End User License Agreement (EULA), pricing, and support terms and choose Accept Offer.
4. Choose Continue to configuration and then choose a Region. You will see a Product ARN displayed, as shown in the following screenshot. This is the model package Amazon Resource Name (ARN) that you need to specify while creating a deployable model using Boto3. Copy the ARN corresponding to your Region and enter it in the following cell.

The code snippets included in this post are sourced from the aws-cohere notebook. If you encounter any issues with this code, refer to the notebook for the most up-to-date version.

!pip install --upgrade cohere-aws
# if you upgrade the package, you need to restart the kernel

from cohere_aws import Client
import boto3

On the Configure for AWS CloudFormation page shown in the following screenshot, under Product Arn, make a note of the last part of the product ARN to use as the value in the variable cohere_package in the following code.

cohere_package = " cohere-rerank-multilingual-v3--13dba038aab73b11b3f0b17fbdb48ea0"

model_package_map = {

"us-east-1": f"arn:aws:sagemaker:us-east-1:865070037744:model-package/{cohere_package}",

"us-east-2": f"arn:aws:sagemaker:us-east-2:057799348421:model-package/{cohere_package}",

"us-west-1": f"arn:aws:sagemaker:us-west-1:382657785993:model-package/{cohere_package}",

"us-west-2": f"arn:aws:sagemaker:us-west-2:594846645681:model-package/{cohere_package}",

"ca-central-1": f"arn:aws:sagemaker:ca-central-1:470592106596:model-package/{cohere_package}",

"eu-central-1": f"arn:aws:sagemaker:eu-central-1:446921602837:model-package/{cohere_package}",

"eu-west-1": f"arn:aws:sagemaker:eu-west-1:985815980388:model-package/{cohere_package}",

"eu-west-2": f"arn:aws:sagemaker:eu-west-2:856760150666:model-package/{cohere_package}",

"eu-west-3": f"arn:aws:sagemaker:eu-west-3:843114510376:model-package/{cohere_package}",

"eu-north-1": f"arn:aws:sagemaker:eu-north-1:136758871317:model-package/{cohere_package}",

"ap-southeast-1": f"arn:aws:sagemaker:ap-southeast-1:192199979996:model-package/{cohere_package}",

"ap-southeast-2": f"arn:aws:sagemaker:ap-southeast-2:666831318237:model-package/{cohere_package}",

"ap-northeast-2": f"arn:aws:sagemaker:ap-northeast-2:745090734665:model-package/{cohere_package}",

"ap-northeast-1": f"arn:aws:sagemaker:ap-northeast-1:977537786026:model-package/{cohere_package}",

"ap-south-1": f"arn:aws:sagemaker:ap-south-1:077584701553:model-package/{cohere_package}",

"sa-east-1": f"arn:aws:sagemaker:sa-east-1:270155090741:model-package/{cohere_package}",

}

region = boto3.Session().region_name

if region not in model_package_map.keys():

raise Exception(f"Current boto3 session region {region} is not supported.")

model_package_arn = model_package_map[region]

Create an endpoint and perform real-time inference

If you want to understand how real-time inference with Amazon SageMaker works, refer to the Amazon SageMaker Developer Guide.

Create an endpoint

To create an endpoint, use the following code.

co = Client(region_name=region)

co.create_endpoint(arn=model_package_arn, endpoint_name="cohere-rerank-multilingual-v3-0", instance_type="ml.g5.2xlarge", n_instances=1)

# If the endpoint is already created, you just need to connect to it

# co.connect_to_endpoint(endpoint_name="cohere-rerank-multilingual-v3-0”)

After the endpoint is created, you can perform real-time inference.

Create the input payload

To create the input payload, use the following code.

documents = [
    {"Title":"Contraseña incorrecta","Content":"Hola, llevo una hora intentando acceder a mi cuenta y sigue diciendo que mi contraseña es incorrecta. ¿Puede ayudarme, por favor?"},
    {"Title":"Confirmation Email Missed","Content":"Hi, I recently purchased a product from your website but I never received a confirmation email. Can you please look into this for me?"},
    {"Title":"أسئلة حول سياسة الإرجاع","Content":"مرحبًا، لدي سؤال حول سياسة إرجاع هذا المنتج. لقد اشتريته قبل بضعة أسابيع وهو معيب"},
    {"Title":"Customer Support is Busy","Content":"Good morning, I have been trying to reach your customer support team for the past week but I keep getting a busy signal. Can you please help me?"},
    {"Title":"Falschen Artikel erhalten","Content":"Hallo, ich habe eine Frage zu meiner letzten Bestellung. Ich habe den falschen Artikel erhalten und muss ihn zurückschicken."},
    {"Title":"Customer Service is Unavailable","Content":"Hello, I have been trying to reach your customer support team for the past hour but I keep getting a busy signal. Can you please help me?"},
    {"Title":"Return Policy for Defective Product","Content":"Hi, I have a question about the return policy for this product. I purchased it a few weeks ago and it is defective."},
    {"Title":"收到错误物品","Content":"早上好，关于我最近的订单，我有一个问题。我收到了错误的商品，需要退货。"},
    {"Title":"Return Defective Product","Content":"Hello, I have a question about the return policy for this product. I purchased it a few weeks ago and it is defective."}
]

Perform real-time inference

To perform real-time inference, use the following code.

response = co.rerank(documents=documents, query='What emails have been about returning items?', rank_fields=["Title","Content"], top_n=5)

Visualize output

To visualize output, use the following code.

print(f'Documents: {response}')

The following screenshot shows the output response.

Cleanup

To avoid any recurring charges, use the following steps to clean up the resources created in this walkthrough.

Delete the model

Now that you have successfully performed a real-time inference, you do not need the endpoint anymore. You can terminate the endpoint to avoid being charged.

co.delete_endpoint()
co.close()

Unsubscribe to the listing (optional)

If you want to unsubscribe to the model package, follow these steps. Before you cancel the subscription, make sure that you don’t have a deployable model created from the model package or using the algorithm. You can find this information by looking at the container name associated with the model.

Steps to unsubscribe from the product from AWS Marketplace:

1. On the Your Software subscriptions page, choose the Machine Learning tab
2. Locate the listing that you want to cancel the subscription for, and then choose Cancel Subscription

Summary

RAG is a capable technique for developing AI applications that integrate real-time data and enable interactive conversations using proprietary information. RAG enhances AI responses by tapping into external, domain-specific knowledge sources, but its effectiveness depends on finding the right source materials. This post focuses on improving search efficiency and accuracy in RAG systems using Cohere Rerank. RAG orchestration typically involves two steps: retrieval of relevant documents and generation of answers. While dense retrieval is efficient for large datasets, it can struggle with complex data and questions due to information compression. Cohere Rerank uses deep learning to evaluate the alignment between documents and queries, outputting a relevance score that enables more nuanced document selection.

Customers can find Cohere Rerank 3 and Cohere Rerank 3 Nimble on Amazon Sagemaker Jumpstart.

About the Authors

Shashi Raina is a Senior Partner Solutions Architect at Amazon Web Services (AWS), where he specializes in supporting generative AI (GenAI) startups. With close to 6 years of experience at AWS, Shashi has developed deep expertise across a range of domains, including DevOps, analytics, and generative AI.

Pradeep Prabhakaran is a Senior Manager – Solutions Architecture at Cohere. In his current role at Cohere, Pradeep acts as a trusted technical advisor to customers and partners, providing guidance and strategies to help them realize the full potential of Cohere’s cutting-edge Generative AI platform.

Read More

Managing cloud costs and understanding resource usage can be a daunting task, especially for organizations with complex AWS deployments. AWS Cost and Usage Reports (AWS CUR) provides valuable data insights, but interpreting and querying the raw data can be challenging.

In this post, we explore a solution that uses generative artificial intelligence (AI) to generate a SQL query from a user’s question in natural language. This solution can simplify the process of querying CUR data stored in an Amazon Athena database using SQL query generation, running the query on Athena, and representing it on a web portal for ease of understanding.

The solution uses Amazon Bedrock, a fully managed service that offers a choice of high-performing foundation models (FMs) from leading AI companies like AI21 Labs, Anthropic, Cohere, Meta, Mistral AI, Stability AI, and Amazon through a single API, along with a broad set of capabilities to build generative AI applications with security, privacy, and responsible AI.

Challenges addressed

The following challenges can hinder organizations from effectively analyzing their CUR data, leading to potential inefficiencies, overspending, and missed opportunities for cost-optimization. We aim to target and simplify them using generative AI with Amazon Bedrock.

• Complexity of SQL queries – Writing SQL queries to extract insights from CUR data can be complex, especially for non-technical users or those unfamiliar with the CUR data structure (unless you’re a seasoned database administrator)
• Data accessibility – To gain insights from structured data in databases, users need to get access to databases, which can be a potential threat to overall data protection
• User-friendliness – Traditional methods of analyzing CUR data often lack a user-friendly interface, making it challenging for non-technical users to take advantage of the valuable insights hidden within the data

Solution overview

The solution that we discuss is a web application (chatbot) that allows you to ask questions related to your AWS costs and usage in natural language. The application generates SQL queries based on the user’s input, runs them against an Athena database containing CUR data, and presents the results in a user-friendly format. The solution combines the power of generative AI, SQL generation, database querying, and an intuitive web interface to provide a seamless experience for analyzing CUR data.

The solution uses the following AWS services:

• Amazon Athena
• Amazon Bedrock
• AWS Billing and Cost Management for cost and usage reports
• Amazon Simple Storage Service (Amazon S3)
• The compute service of your choice on AWS to call Amazon Bedrock APIs. This could be Amazon Elastic Compute Cloud (Amazon EC2), AWS Lambda, AWS SDK, Amazon SageMaker notebooks, or your workstation if you are doing a quick proof of concept. For the purpose of this post, this code is running on a t3a.micro EC2 instance with Amazon Linux 2023.

 The following diagram illustrates the solution architecture.

Figure 1. Architecture of Solution

The data flow consists of the following steps:

1. The CUR data is stored in Amazon S3.
2. Athena is configured to access and query the CUR data stored in Amazon S3.
3. The user interacts with the Streamlit web application and submits a natural language question related to AWS costs and usage.

Figure 2. Shows the Chatbot Dashboard to ask question

4. The Streamlit application sends the user’s input to Amazon Bedrock, and the LangChain application facilitates the overall orchestration.
5. The LangChain code uses the BedrockChat class from LangChain to invoke the FM and interact with Amazon Bedrock to generate a SQL query based on the user’s input.

Figure 3. Shows initialization of SQL chain

6. The generated SQL query is run against the Athena database using the FM on Amazon Bedrock, which queries the CUR data stored in Amazon S3.
7. The query results are returned to the LangChain application.

Figure 4. Shows generated Query in the application output logs

8. LangChain sends the SQL query and query results back to the Streamlit application.
9. The Streamlit application displays the SQL query and query results to the user in a formatted and user-friendly manner.

Figure 5. Shows final output presented on the chat bot webapp including SQL Query and the Query results

Prerequisites

To set up this solution, you should have the following prerequisites:

• An AWS account with access to AWS Cost Explorer, Athena, and Amazon S3. (This is a proof of concept setup. You should follow the least privilege model when using AWS Identity and Access Management (IAM), refer to the IAM security best practices documentation, and conduct your own due diligence when setting this up.)
• CUR data stored in an S3 bucket. For instructions, see Creating Cost and Usage Reports.
• Athena set up to analyze the data from your S3 bucket holding CUR data. For instructions, see Querying Cost and Usage Reports using Amazon Athena.
• An AWS compute environment created to host the code and call the Amazon Bedrock APIs.
• An AWS Identity and Access Management (IAM) role with permission to access Amazon Bedrock and access to FMs in Amazon Bedrock.

Configure the solution

Complete the following steps to set up the solution:

1. Create an Athena database and table to store your CUR data. Make sure the necessary permissions and configurations are in place for Athena to access the CUR data stored in Amazon S3.
2. Set up your compute environment to call Amazon Bedrock APIs. Make sure you associate an IAM role with this environment that has IAM policies that grant access to Amazon Bedrock.
3. When your instance is up and running, install the following libraries that are used for working within the environment:

pip install langchain==0.2.0 langchain-experimental==0.0.59 langchain-community==0.2.0 langchain-aws==0.1.4 pyathena==3.8.2 sqlalchemy==2.0.30 streamlit==1.34.0

4. Use the following code to establish a connection to the Athena database using the langchain library and the pyathena Configure the language model to generate SQL queries based on user input using Amazon Bedrock. You can save this file as cur_lib.py.

from langchain_experimental.sql import SQLDatabaseChain
from langchain_community.utilities import SQLDatabase
from sqlalchemy import create_engine, URL
from langchain_aws import ChatBedrock as BedrockChat
from pyathena.sqlalchemy.rest import AthenaRestDialect

class CustomAthenaRestDialect(AthenaRestDialect):
    def import_dbapi(self):
        import pyathena
        return pyathena

# DB Variables
connathena = "athena.us-west-2.amazonaws.com"
portathena = '443'
schemaathena = 'mycur'
s3stagingathena = 's3://cur-data-test01/athena-query-result/'
wkgrpathena = 'primary'
connection_string = f"awsathena+rest://@{connathena}:{portathena}/{schemaathena}?s3_staging_dir={s3stagingathena}/&work_group={wkgrpathena}"
url = URL.create("awsathena+rest", query={"s3_staging_dir": s3stagingathena, "work_group": wkgrpathena})
engine_athena = create_engine(url, dialect=CustomAthenaRestDialect(), echo=False)
db = SQLDatabase(engine_athena)

# Setup LLM
model_kwargs = {"temperature": 0, "top_k": 250, "top_p": 1, "stop_sequences": ["nnHuman:"]}
llm = BedrockChat(model_id="anthropic.claude-3-sonnet-20240229-v1:0", model_kwargs=model_kwargs)

# Create the prompt
QUERY = """
Create a syntactically correct athena query for AWS Cost and Usage report to run on the my_c_u_r table in mycur database based on the question, then look at the results of the query and return the answer as SQLResult like a human
{question}
"""
db_chain = SQLDatabaseChain.from_llm(llm, db, verbose=True)

def get_response(user_input):
    question = QUERY.format(question=user_input)
    result = db_chain.invoke(question)
    query = result["result"].split("SQLQuery:")[1].strip()
    rows = db.run(query)
    return f"SQLQuery: {query}nSQLResult: {rows}"

5. Create a Streamlit web application to provide a UI for interacting with the LangChain application. Include the input fields for users to enter their natural language questions and display the generated SQL queries and query results. You can name this file cur_app.py.

import streamlit as st
from cur_lib import get_response
import os

st.set_page_config(page_title="AWS Cost and Usage Chatbot", page_icon="chart_with_upwards_trend", layout="centered", initial_sidebar_state="auto",
menu_items={
        'Get Help': 'https://docs.aws.amazon.com/cur/latest/userguide/cur-create.html',
        #'Report a bug':,
        'About': "# The purpose of this app is to help you get better understanding of your AWS Cost and Usage report!"
    })#HTML title
st.title("_:orange[Simplify] CUR data_ :sunglasses:")

def format_result(result):
    parts = result.split("nSQLResult: ")
    if len(parts) > 1:
        sql_query = parts[0].replace("SQLQuery: ", "")
        sql_result = parts[1].strip("[]").split("), (")
        formatted_result = []
        for row in sql_result:
            formatted_result.append(tuple(item.strip("(),'") for item in row.split(", ")))
        return sql_query, formatted_result
    else:
        return result, []

def main():
    # Get the current directory
    current_dir = os.path.dirname(os.path.abspath(__file__))
    st.markdown("<div class='main'>", unsafe_allow_html=True)
    st.title("AWS Cost and Usage chatbot")
    st.write("Ask a question about your AWS Cost and Usage Report:")

6. Connect the LangChain application and Streamlit web application by calling the get_response Format and display the SQL query and result in the Streamlit web application. Append the following code with the preceding application code:

# Create a session state variable to store the chat history
    if "chat_history" not in st.session_state:
        st.session_state.chat_history = []

    user_input = st.text_input("You:", key="user_input")

    if user_input:
        try:
            result = get_response(user_input)
            sql_query, sql_result = format_result(result)
            st.code(sql_query, language="sql")
            if sql_result:
                st.write("SQLResult:")
                st.table(sql_result)
            else:
                st.write(result)
            st.session_state.chat_history.append({"user": user_input, "bot": result})
            st.text_area("Conversation:", value="n".join([f"You: {chat['user']}nBot: {chat['bot']}" for chat in st.session_state.chat_history]), height=300)
        except Exception as e:
            st.error(str(e))

    st.markdown("</div>", unsafe_allow_html=True)

if __name__ == "__main__":
    main()

7. Deploy the Streamlit application and LangChain application to your hosting environment, such as Amazon EC2, or a Lambda function.

Clean up

Unless you invoke Amazon Bedrock with this solution, you won’t incur charges for it. To avoid ongoing charges for Amazon S3 storage for saving the CUR reports, you can remove the CUR data and S3 bucket. If you set up the solution using Amazon EC2, make sure you stop or delete the instance when you’re done.

Benefits

This solution offers the following benefits:

• Simplified data analysis – You can analyze CUR data using natural language using generative AI, eliminating the need for advanced SQL knowledge
• Increased accessibility – The web-based interface makes it efficient for non-technical users to access and gain insights from CUR data without needing credentials for the database
• Time-saving – You can quickly get answers to your cost and usage questions without manually writing complex SQL queries
• Enhanced visibility – The solution provides visibility into AWS costs and usage, enabling better cost-optimization and resource management decisions

Summary

The AWS CUR chatbot solution uses Anthropic Claude on Amazon Bedrock to generate SQL queries, database querying, and a user-friendly web interface to simplify the analysis of CUR data. By allowing you to ask natural language questions, the solution removes barriers and empowers both technical and non-technical users to gain valuable insights into AWS costs and resource usage. With this solution, organizations can make more informed decisions, optimize their cloud spending, and improve overall resource utilization. We recommend that you do due diligence while setting this up, especially for production; you can choose other programming languages and frameworks to set it up according to your preference and needs.

Amazon Bedrock enables you to build powerful generative AI applications with ease. Accelerate your journey by following the quick start guide on GitHub and using Amazon Bedrock Knowledge Bases to rapidly develop cutting-edge Retrieval Augmented Generation (RAG) solutions or enable generative AI applications to run multistep tasks across company systems and data sources using Amazon Bedrock Agents.

About the Author

Anutosh is a Solutions Architect at AWS India. He loves to dive deep into his customers’ use cases to help them navigate through their journey on AWS. He enjoys building solutions in the cloud to help customers. He is passionate about migration and modernization, data analytics, resilience, cybersecurity, and machine learning.

Read More

Intricate workflows that require dynamic and complex API orchestration can often be complex to manage. In industries like insurance, where unpredictable scenarios are the norm, traditional automation falls short, leading to inefficiencies and missed opportunities. With the power of intelligent agents, you can simplify these challenges. In this post, we explore how chaining domain-specific agents using Amazon Bedrock Agents can transform a system of complex API interactions into streamlined, adaptive workflows, empowering your business to operate with agility and precision.

Amazon Bedrock is a fully managed service that offers a choice of high-performing foundation models (FMs) from leading artificial intelligence (AI) companies like AI21 Labs, Anthropic, Cohere, Meta, Mistral AI, Stability AI, and Amazon through a single API, along with a broad set of capabilities to build generative AI applications with security, privacy, and responsible AI.

Benefits of chaining Amazon Bedrock Agents

Designing agents is like designing other software components—they tend to work best when they have a focused purpose. When you have focused, single-purpose agents, combining them into chains can allow them to solve significantly complex problems together. Using natural language processing (NLP) and OpenAPI specs, Amazon Bedrock Agents dynamically manages API sequences, minimizing dependency management complexities. Additionally, agents enable conversational context management in real-time scenarios, using session IDs and, if necessary, backend databases like Amazon DynamoDB for extended context storage. By using prompt instructions and API descriptions, agents collect essential information from API schemas to solve specific problems efficiently. This approach not only enhances agility and flexibility, but also demonstrates the value of chaining agents to simplify complex workflows and solve larger problems effectively.

In this post, we explore an insurance claims use case, where we demonstrate the concept of chaining with Amazon Bedrock Agents. This involves an orchestrator agent calling and interacting with other agents to collaboratively perform a series of tasks, enabling efficient workflow management.

Solution overview

For our use case, we develop a workflow for an insurance digital assistant focused on streamlining tasks such as filing claims, assessing damages, and handling policy inquiries. The workflow simulates API sequencing dependencies, such as conducting fraud checks during claim creation and analyzing uploaded images for damage assessment if the user provides images. The orchestration dynamically adapts to user scenarios, guided by natural language prompts from domain-specific agents like an insurance orchestrator agent, policy information agent, and damage analysis notification agent. Using OpenAPI specifications and natural language prompts, the API sequencing in our insurance digital assistant adapts to dynamic user scenarios, such as users opting in or out of image uploads for damage assessment, failing fraud checks or choosing to ask a variety of questions related to their insurance policies and coverages. This flexibility is achieved by chaining domain-specific agents like the insurance orchestrator agent, policy information agent, and damage analysis notification agent.

Traditionally, insurance processes are rigid, with fixed steps for tasks like fraud detection. However, agent chaining allows for greater flexibility and adaptability, enabling the system to respond to real-time user inputs and variations in scenarios. For instance, instead of strictly adhering to predefined thresholds for fraud checks, the agents can dynamically adjust the workflow based on user interactions and context. Similarly, when users choose to upload images while filing a claim, the workflow can perform real-time damage analysis and immediately send a summary to claims adjusters for further review. This enables a quicker response and more accurate decision-making. This approach not only streamlines the claims process but also allows for a more nuanced and efficient handling of tasks, providing the necessary balance between automation and human intervention. By chaining Amazon Bedrock Agents, we create a system that is adaptable. This system caters to diverse user needs while maintaining the integrity of business processes.

The following diagram illustrates the end-to-end insurance claims workflow using chaining with Amazon Bedrock Agents.

The diagram shows how specialized agents use various tools to streamline the entire claims process—from filing claims and assessing damages to answering customer questions about insurance policies.

Prerequisites

Before proceeding, make sure you have the following resources set up:

• An AWS account. If you don’t have an account, you can sign up for one.
• Access as an AWS Identity and Access Management (IAM) administrator or an IAM user that has permissions for:
  • Deploying AWS CloudFormation
  • Creating and managing Amazon Simple Storage Service (Amazon S3) buckets and uploading objects.
  • Creating and updating Amazon Simple Queue Service (Amazon SQS) queues, AWS Lambda functions, and Amazon API Gateway.
  • Creating and managing IAM roles.
  • Access to Amazon Bedrock, Anthropic’s Claude models on Amazon Bedrock, and the Cohere Embed English embedding model on Amazon Bedrock. You must explicitly enable access to models before they can be used with Amazon Bedrock. For instructions, refer to Model access.

Deploy the solution with AWS CloudFormation

Complete the following steps to set up the solution resources:

1. Sign in to the AWS Management Console as an IAM administrator or appropriate IAM user.
2. Choose Launch Stack to deploy the CloudFormation template.
   
3. Provide the necessary parameters and create the stack.

For this setup, we use us-east-1 as our AWS Region, the Anthropic Claude 3 Haiku model for orchestrating the flow between the different agents, the Anthropic Claude 3 Sonnet model for damage analysis of the uploaded images, and the Cohere Embed English V3 model as an embedding model to translate text from the insurance policy documents into numerical vectors, which allows for efficient search, comparison, and categorization of the documents.

If you want to choose other models on Amazon Bedrock, you can do so by making appropriate changes in the CloudFormation template. Check for appropriate model support in the Region and the features that are supported by the models.

This will take about 15 minutes to deploy the solution. After the stack is deployed, you can view the various outputs of the CloudFormation stack on the Outputs tab, as shown in the following screenshot.

The following screenshot shows the three Amazon Bedrock agents that were deployed in your account.

Test the claims creation, damage detection, and notification workflows

The first part of the deployed solution is to mimic filing a new insurance claim, fraud detection, optional damage analysis of uploading images, and subsequent notification to claims adjusters. This is a smaller version of task automation to fulfill a particular business problem achieved by chaining agents, each performing a set of specific tasks. The agents work in harmony to solve the larger function of insurance claims handling.

Let’s explore the architecture of the claim creation workflow, where the insurance orchestrator agent and the damage analysis notification agent work together to simulate filing new claims, assessing damages, and sending a summary of damages to the claim adjusters for human oversight. The following diagram illustrates this workflow.

In this workflow, the insurance orchestrator agent mimics fraud detection and claims creation as well as orchestrates handing off the responsibility to other task-specific agents. The image damage analysis notification agent is responsible for doing a preliminary analysis of the images uploaded for a damage. This agent invokes a Lambda function that internally calls the Anthropic Claude Sonnet large language model (LLM) on Amazon Bedrock to perform preliminary analysis on the images. The LLM generates a summary of the damage, which is sent to an SQS queue, and is subsequently reviewed by the claim adjusters.

The NLP instruction prompts combined with the OpenAPI specifications for each action group guide the agents in their decision-making process, determining which action group to invoke, the sequence of invocation, and the required parameters for calling specific APIs.

Use the UI to invoke the claims processing workflow

Complete the following steps to invoke the claims processing workflow:

1. From the outputs of the CloudFormation stack, choose the URL for HttpApiEndpoint.

2. You can ask the chatbots sample questions to start exploring the functionality of filing a new claim.

In the following example, we ask for filing a new claim and uploading images as evidence for the claim.

3. On the Amazon SQS console, you can view the SQS queue that has been created by the CloudFormation stack and check the message that shows the damage analysis from the image performed by our LLM.

Test the policy information workflow

The following diagram shows the architecture of just the policy information agent. The policy agent accesses the Policy Information API to extract answers to insurance-related questions from unstructured policy documents such as PDF files.

The policy information agent is responsible for doing a lookup against the insurance policy documents stored in the knowledge base. The agent invokes a Lambda function that will internally invoke the knowledge base to find answers to policy-related questions.

Set up the policy documents and metadata in the data source for the knowledge base

We use Amazon Bedrock Knowledge Bases to manage our documents and metadata. As part of deploying the solution, the CloudFormation stack created a knowledge base. Complete the following steps to set up its data source:

1. On the Amazon Bedrock console, navigate to the deployed knowledge base and navigate to the S3 bucket that is mentioned as its data source.

2. Upload a few insurance policy documents and metadata documents to the S3 bucket to mimic the naming conventions as shown in the following screenshot.

The naming conventions are <Type of Policy>_PolicyNumber.pdf for the insurance policy PDF documents and <Type of Policy>_PolicyNumber.pdf.metadata.json for the metadata documents.

The following screenshot shows an example of what a sample metadata.json file looks like.

3. After the documents are uploaded to Amazon S3, navigate to the deployed knowledge base, select the data source, and choose Sync.

To understand more about how metadata support in Knowledge Bases on Amazon Bedrock helps you get accurate results, refer to Amazon Bedrock Knowledge Bases now supports metadata filtering to improve retrieval accuracy.

4. Now you can go back to the UI and start asking questions related to the policy documents.

The following screenshot shows the set of questions we asked for finding answers related to policy coverage.

Clean up

To avoid unexpected charges, complete the following steps to clean up your resources:

1. Delete the contents from the S3 buckets corresponding to the ImageBucketName and PolicyDocumentsBucketName keys from the outputs of the CloudFormation stack.
2. Delete the deployed stack using the AWS CloudFormation console.

Best practices

The following are some additional best practices that you can follow for your agents:

• Automated testing – Implement automated tests using tools to regularly test the orchestration workflows. You can use mock APIs to simulate various scenarios and validate the agent’s decision-making process.
• Version control – Maintain version control for your agent configurations and prompts in a repository. This provides traceability and quick rollback if needed.
• Monitoring and logging – Use Amazon CloudWatch to monitor agent interactions and API calls. Set up alarms for unexpected behaviors or failures.
• Continuous integration – Set up a continuous integration and delivery (CI/CD) pipeline that integrates automated testing, prompt validation, and deployment to maintain smooth updates without disrupting ongoing workflows.

Conclusion

In this post, we demonstrated the power of chaining Amazon Bedrock agents, offering a fresh perspective on integrating back-office automation workflows and enterprise APIs. This solution offers several benefits: as new enterprise APIs emerge, dependencies in existing ones can be minimized, reducing coupling. Moreover, Amazon Bedrock Agents can maintain conversational context, enabling follow-up queries to use conversation history. For extended contextual memory, a more persistent backend implementation can be considered.

To learn more, refer to Amazon Bedrock Agents.

About the Author

Piyali Kamra is a seasoned enterprise architect and a hands-on technologist who has over two decades of experience building and executing large scale enterprise IT projects across geographies. She believes that building large scale enterprise systems is not an exact science but more like an art, where you can’t always choose the best technology that comes to one’s mind but rather tools and technologies must be carefully selected based on the team’s culture , strengths, weaknesses and risks, in tandem with having a futuristic vision as to how you want to shape your product a few years down the road.

Read More

Amazon SageMaker is a fully managed machine learning (ML) service. With SageMaker, data scientists and developers can quickly and confidently build, train, and deploy ML models into a production-ready hosted environment. SageMaker provides a broad selection of ML infrastructure and model deployment options to help meet your ML inference needs. It also helps scale your model deployment, manage models more effectively in production, and reduce operational burden.

Although early large language models (LLMs) were limited to processing text inputs, the rapid evolution of these AI systems has enabled LLMs to expand their capabilities to handle a wide range of media types, including images, video, and audio, ushering in the era of multimodal models. Multimodal is a type of deep learning using multiple modalities of data, such as text, audio, or images. Multimodal inference adds challenges of large data transfer overhead and slow response times. For instance, in a typical chatbot scenario, users initiate the conversation by providing a multimedia file or a link as input payload, followed by a back-and-forth dialogue, asking questions or seeking information related to the initial input. However, transmitting large multimedia files with every request to a model inference endpoint can significantly impact the response times and latency, leading to an unsatisfactory user experience. For example, sending a 500 MB input file could potentially add 3–5 seconds to the response time, which is unacceptable for a chatbot aiming to deliver a seamless and responsive interaction.

We are announcing the availability of sticky session routing on Amazon SageMaker Inference which helps customers improve the performance and user experience of their generative AI applications by leveraging their previously processed information. Amazon SageMaker makes it easier to deploy ML models including foundation models (FMs) to make inference requests at the best price performance for any use case.

By enabling sticky sessions routing, all requests from the same session are routed to the same instance, allowing your ML application to reuse previously processed information to reduce latency and improve user experience. This is particularly valuable when you want to use large data payloads or need seamless interactive experiences. By using your previous inference requests, you can now take advantage of this feature to build innovative state-aware AI applications on SageMaker. To do, you create a session ID with your first request, and then use that session ID to indicate that SageMaker should route all subsequent requests to the same instance. Sessions can also be deleted when done to free up resources for new sessions.

This feature is available in all AWS Regions where SageMaker is available. To learn more about deploying models on SageMaker, see Amazon SageMaker Model Deployment. For more about this feature, refer to Stateful sessions with Amazon SageMaker models.

Solution overview

SageMaker simplifies the deployment of models, enabling chatbots and other applications to use their multimodal capabilities with ease. SageMaker has implemented a robust solution that combines two key strategies: sticky session routing in SageMaker with load balancing, and stateful sessions in TorchServe. Sticky session routing makes sure all requests from a user session are serviced by the same SageMaker server instance. Stateful sessions in TorchServe cache the multimedia data in GPU memory from the session start request and minimize loading and unloading of this data from GPU memory for improved response times.

With this focus on minimizing data transfer overhead and improving response time, our approach makes sure the initial multimedia file is loaded and processed only one time, and subsequent requests within the same session can use the cached data.

Let’s look at the sequence of events when a client initiates a sticky session on SageMaker:

1. In the first request, you call the Boto3 SageMaker runtime invoke_endpoint with session-id=NEW_SESSION in the header and a payload indicating an open session type of request. SageMaker then creates a new session and stores the session ID. The router initiates an open session (this API is defined by the client; it could be some other name like start_session) with the model server, in this case TorchServe, and responds back with 200 OK along with the session ID and time to live (TTL), which is sent back to the client.
   

2. Whenever you need to use the same session to perform subsequent actions, you pass the session ID as part of the invoke_endpoint call, which allows SageMaker to route all the subsequent requests to the same model server instance.
3. To close or delete a session, you can use invoke_endpoint with a payload indicating a close session type of request along with the session ID. The SageMaker router first checks if the session exists. If it does, the router initiates a close session call to the model server, which responds back with a successful 200 OK along with session ID, which is sent back to the client. In the scenario, when the session ID doesn’t exist, the router responds back with a 400 response.
   

In the following sections, we walk through an example of how you can use sticky routing in SageMaker to achieve stateful model inference. For this post, we use the LLaVA: Large Language and Vision Assistant model. LLaVa is a multimodal model that accepts images and text prompts.

We use LLaVa to upload an image and then ask questions about the image without having to resend the image for every request. The image is cached in the GPU memory as opposed to the CPU memory, so we don’t have to incur the latency cost of moving this image from CPU memory to GPU memory on every call.

We use TorchServe as our model server for this example. TorchServe is a performant, flexible and easy to use tool for serving PyTorch models in production. TorchServe supports a wide array of advanced features, including dynamic batching, microbatching, model A/B testing, streaming, torch XLA, tensorRT, ONNX and IPEX. Moreover, it seamlessly integrates PyTorch’s large model solution, PiPPy, enabling efficient handling of large models. Additionally, TorchServe extends its support to popular open-source libraries like DeepSpeed, Accelerate, Fast Transformers, and more, expanding its capabilities even further.

The following are the main steps to deploy the LLava model. The section below introduces the steps conceptually, so you’ll have a better grasp of the overall deployment workflow before diving into the practical implementation details in the subsequent section.

Build a TorchServe Docker container and push it to Amazon ECR

The first step is to build a TorchServe Docker container and push it to Amazon Elastic Container Registry (Amazon ECR). Because we’re using a custom model, we use the bring your own container approach. We use one of the AWS provided deep learning containers as our base, namely pytorch-inference:2.3.0-gpu-py311-cu121-ubuntu20.04-sagemaker.

Build TorchServe model artifacts and upload them to Amazon S3

We use torch-model-archiver to gather all the artifacts, like custom handlers, the LlaVa model code, the data types for request and response, model configuration, prediction API, and other utilities. Then we upload the model artifacts to Amazon Simple Storage Service (Amazon S3).

Create the SageMaker endpoint

To create the SageMaker endpoint, complete the following steps:

1. To create the model, use the SageMaker Python SDK Model class and as inputs. Specify the S3 bucket you created earlier to upload the TorchServe model artifacts and the image_uri of the Docker container you created.

SageMaker expects the session ID in X-Amzn-SageMaker-Session-Id format; you can specify that in the environment properties to the model.

2. To deploy the model and create the endpoint, specify the initial instance count to match the load, instance type, and timeouts.
3. Lastly, create a SageMaker Python SDK Predictor by passing in the endpoint name.

Run inference

Complete the following steps to run inference:

1. Use an open session to send a URL to the image you want to ask questions about.

This is a custom API we have defined for our use case (see inference_api.py). You can define the inputs, outputs, and APIs to suit your business use case. For this use case, we use an open session to send a URL to the image we want to ask questions about. For the session ID header value, use the special string NEW_SESSION to indicate this is the start of a session. The custom handler you wrote downloads the image, converts it to a tensor, and caches that in the GPU memory. We do this because we have access to the LLaVa source code; we could also modify the original predict.py file from LLaVa model to accept a tensor instead of a PIL image. By caching the tensor in GPU, we have saved some inference time by not moving the image from CPU memory to GPU memory for every call. If you don’t have access to the model source code, you have to cache the image in CPU memory. Refer to inference_api.py for this source code. The open session API call returns a session ID, which you use for the rest of the calls in this session.

2. To send a text prompt, get the session ID from the open session and send it along with the text prompt.

inference_api.py looks up the cache in GPU for the image based on the session ID and uses that for inference. This returns the LLaVa model output as a string.

3. Repeat the previous step to send a different text prompt.
4. When you’re done with all the text prompts, use the session ID to close the session.

In inference_api.py, we no longer hold on to the image cache in GPU.

The source code for this example is in the GitHub repo. You can run the steps using the following notebook.

Prerequisites

Use the following code to deploy an AWS CloudFormation stack that creates an AWS Identity and Access Management (IAM) role to deploy the SageMaker endpoints:

aws cloudformation create-stack --stack-name sm-stateful-role 
--template-body https://raw.githubusercontent.com/aws-samples/sagemaker-genai-hosting-examples/main/LLava/torchserve/workspace/sm_role.yaml 
--capabilities CAPABILITY_NAMED_IAM 
--region us-west-2

Create a SageMaker notebook instance

Complete the following steps to create a notebook instance for LLaVa model deployment:

1. On the SageMaker console, choose Notebooks in the navigation pane.
   

2. Choose Create notebook instance.
   

3. In the Notebook instance settings section, under Additional configuration, choose at least 500 GB for the storage volume.
   

4. In the Permissions and encryption section, choose to use an existing IAM role, and choose the role you created in the prerequisites (sm-stateful-role-xxx).

You can get the full name of the role on the AWS CloudFormation console, on the Resources tab of the stack sm-stateful-role.

5. In the Git repositories section, for Git repository URL, enter https://github.com/aws-samples/sagemaker-genai-hosting-examples.git.
   

6. Choose Create notebook instance.

Run the notebook

When the notebook is ready, complete the following steps:

1. On the SageMaker console, choose Notebooks in the navigation pane.
2. Choose Open JupyterLab for this new instance.
   

3. In JupyterLab, navigate to LLava using the file explorer.
   

4. Navigate to torchserve /workspace / and open the notebook llava_stateful_deploy_infer.ipynb.
   

5. Run the notebook.

The ./build_and_push.sh script takes approximately 30 minutes to run. You can also run the ./build_and_push.sh script in a terminal for better feedback. Note the input parameters from the previous step and make sure you’re in the right directory (sagemaker-genai-hosting-examples/LLava/torchserve/workspace).

The model.deploy() step also takes 20–30 minutes to complete.

6. When you’re done, run the last cleanup cell.
   

7. Additionally, delete the SageMaker notebook instance.

Troubleshooting

When you run ./build_and_push.sh, you might get the following error:

./build_and_push.sh: line 48: docker: command not found

This means you’re not using SageMaker notebooks, and are probably using Amazon SageMaker Studio. Docker is not installed in SageMaker Studio by default.

Look at the screen shot below to learn how to open Amazon SageMaker Notebook.

Conclusion

In this post, we explained how the new sticky routing feature in Amazon SageMaker allows you to achieve ultra-low latency and enhance your end-user experience when serving multi-modal models. You can use the provided notebook and create stateful endpoints for your multimodal models to enhance your end-user experience.

Try out this solution for your own use case, and let us know your feedback and questions in the comments.

About the authors

Harish Rao is a senior solutions architect at AWS, specializing in large-scale distributed AI training and inference. He empowers customers to harness the power of AI to drive innovation and solve complex challenges. Outside of work, Harish embraces an active lifestyle, enjoying the tranquility of hiking, the intensity of racquetball, and the mental clarity of mindfulness practices.

Raghu Ramesha is a Senior GenAI/ML Solutions Architect on the Amazon SageMaker Service team. He focuses on helping customers build, deploy, and migrate ML production workloads to SageMaker at scale. He specializes in machine learning, AI, and computer vision domains, and holds a master’s degree in computer science from UT Dallas. In his free time, he enjoys traveling and photography.

Lingran Xia is a software development engineer at AWS. He currently focuses on improving inference performance of machine learning models. In his free time, he enjoys traveling and skiing.

Naman Nandan is a software development engineer at AWS, specializing in enabling large scale AI/ML inference workloads on SageMaker using TorchServe, a project jointly developed by AWS and Meta. In his free time, he enjoys playing tennis and going on hikes.

Li Ning is a senior software engineer at AWS with a specialization in building large-scale AI solutions. As a tech lead for TorchServe, a project jointly developed by AWS and Meta, her passion lies in leveraging PyTorch and AWS SageMaker to help customers embrace AI for the greater good. Outside of her professional endeavors, Li enjoys swimming, traveling, following the latest advancements in technology, and spending quality time with her family.

Frank Liu is a Principal Software Engineer for AWS Deep Learning. He focuses on building innovative deep learning tools for software engineers and scientists. Frank has in-depth knowledge on the infrastructure optimization and Deep Learning acceleration.

Deepika Damojipurapu is a Senior Technical Account Manager at AWS, specializing in distributed AI training and inference. She helps customers unlock the full potential of AWS by providing consultative guidance on architecture and operations, tailored to their specific applications and use cases. When not immersed in her professional responsibilities, Deepika finds joy in spending quality time with her family – exploring outdoors, traveling to new destinations, cooking wholesome meals together, creating cherished memories.

Alan Tan is a Principal Product Manager with SageMaker, leading efforts on large model inference. He’s passionate about applying machine learning to building novel solutions. Outside of work, he enjoys the outdoors.

Read More