Monthly Archives: October 2024

Global Resiliency is a new Amazon Lex capability that enables near real-time replication of your Amazon Lex V2 bots in a second AWS Region. When you activate this feature, all resources, versions, and aliases associated after activation will be synchronized across the chosen Regions. With Global Resiliency, the replicated bot resources and aliases in the second Region will have the same identifiers as those in the source Region. This consistency allows you to seamlessly route traffic to any Region by simply changing the Region identifier, providing uninterrupted service availability. In the event of a Regional outage or disruption, you can swiftly redirect your bot traffic to a different Region. Applications now have the ability to use replicated Amazon Lex bots across Regions in an active-active or active-passive manner for improved availability and resiliency. With Global Resiliency, you no longer need to manually manage separate bots across Regions, because the feature automatically replicates and keeps Regional configurations in sync. With just a few clicks or commands, you gain robust Amazon Lex bot replication capabilities. Applications that are using Amazon Lex bots can now fail over from an impaired Region seamlessly, minimizing the risk of costly downtime and maintaining business continuity. This feature streamlines the process of maintaining robust and highly available conversational applications. These include interactive voice response (IVR) systems, chatbots for digital channels, and messaging platforms, providing a seamless and resilient customer experience.

In this post, we walk you through enabling Global Resiliency for a sample Amazon Lex V2 bot. We showcase the replication process of bot versions and aliases across multiple Regions. Additionally, we discuss how to handle integrations with AWS Lambda and Amazon CloudWatch after enabling Global Resiliency.

Solution overview

For this exercise, we create a BookHotel bot as our sample bot. We use an AWS CloudFormation template to build this bot, including defining intents, slots, and other required components such as a version and alias. Throughout our demonstration, we use the us-east-1 Region as the source Region, and we replicate the bot in the us-west-2 Region, which serves as the replica Region. We then replicate this bot, enable logging, and integrate it with a Lambda function.

To better understand the solution, refer to the following architecture diagram.

1. Enabling Global Resiliency for an Amazon Lex bot is straightforward using the AWS Management Console, AWS Command Line Interface (AWS CLI), or APIs. We walk through the instructions to replicate the bot later in this post.
2. After replication is successfully enabled, the bot will be replicated across Regions, providing a unified experience. This allows you to distribute IVR or chat application requests between Regions in either an active-active or active-passive setup, depending on your use case.
3. A key benefit of Global Resiliency is that developers can continuously work on bot improvements in the source Region, and changes are automatically synchronized to the replica Region. This streamlines the development workflow without compromising resiliency.

At the time of writing, Global Resiliency only works with predetermined pairs of Regions. For more information, see Use Global Resiliency to deploy bots to other Regions.

Prerequisites

You should have the following prerequisites:

• An AWS account with administrator access
• Access to Amazon Lex Global Resiliency (contact your Amazon Connect Solutions Architect or Technical Account Manager)
• Working knowledge of the following services:
  • AWS CloudFormation
  • Amazon CloudWatch
  • AWS Lambda
  • Amazon Lex

Create a sample Amazon Lex bot

To set up a sample bot for our use case, refer to Manage your Amazon Lex bot via AWS CloudFormation templates. For this example, we create a bot named BookHotel in the source Region (us-east-1). Complete the following steps:

1. Download the CloudFormation template and deploy it in the source Region (us-east-1). For instructions, see Create a stack from the CloudFormation console.

Upon successful deployment, the BookHotel bot will be created in the source Region.

2. On the Amazon Lex console, choose Bots in the navigation pane and locate the BookHotel.

Verify that the Global Resiliency option is available under Deployment in the navigation pane. If this option isn’t visible, the Global Resiliency feature may not be enabled for your account. In this case, refer to the prerequisites section for enabling the Global Resiliency feature.

Our sample BookHotel bot has one version (Version 1, in addition to the draft version) and an alias named BookHotelDemoAlias (in addition to the TestBotAlias).

Enable Global Resiliency

To activate Global Resiliency and set up bot replication in a replica Region, complete the following steps:

1. On the Amazon Lex console, choose us-east-1 as your Region.
2. Choose Bots in the navigation pane and locate the BookHotel.
3. Under Deployment in the navigation pane, choose Global Resiliency.

You can see the replication details here. Because you haven’t enabled Global Resiliency yet, all the details are blank.

4. Choose Create replica to create a draft version of your bot.

In your source Region (us-east-1), after the bot replication is complete, you will see Replication status as Enabled.

5. Switch to the replica Region (us-west-2).

You can see that the BookHotel bot is replicated. This is a read-only replica and the bot ID in the replica Region matches the bot ID in the source Region.

6. Under Deployment in the navigation pane, choose Global Resiliency.

You can see the replication details here, which are the same as that in the source Region BookHotel bot.

You have verified that the bot is replicated successfully after Global Resiliency is enabled. Only new versions and aliases created from this point onward will be replicated. As a next step, we create a bot version and alias to demonstrate the replication.

Create a new bot version and alias

Complete the following steps to create a new bot version and alias:

1. On the Amazon Lex console in your source Region (us-east-1), navigate to the BookHotel.
2. Choose Bot versions in the navigation pane, and choose Create new version to create Version 2.

Version 2 now has Global Resiliency enabled, whereas Version 1 and the draft version do not, because they were created prior to enabling Global Resiliency.

3. Choose Aliases in the navigation pane, then choose Create new alias.
4. Create a new alias for the BookHotel bot called BookHotelDemoAlias_GR and point that to the new version.

Similarly, the BookHotelDemoAlias_GR now has Global Resiliency enabled, whereas aliases created before enabling Global Resiliency, such as BookHotelDemoAlias and TestBotAlias, don’t have Global Resiliency enabled.

5. Choose Global Resiliency in the navigation pane to view the source and replication details.

The details for Last replicated version are now updated to Version 2.

6. Switch to the replica Region (us-west-2) and choose Global Resiliency in the navigation pane.

You can see that the new Global Resiliency enabled version (Version 2) is replicated and the new alias BookHotelDemoAlias_GR is also present.

You have verified that the new version and alias were created after Global Resiliency is replicated to the replica Region. You can now make Amazon Lex runtime calls to both Regions.

Handling integrations with Lambda and CloudWatch after enabling Global Resiliency

Amazon Lex has integrations with other AWS services such as enabling custom logic with Lambda functions and logging with conversation logs using CloudWatch and Amazon Simple Storage Service (Amazon S3). In this section, we associate a Lambda function and CloudWatch group for the BookHotel bot in the source Region (us-east-1) and validate its association in the replica Region (us-west-2).

1. Download the CloudFormation template to deploy a sample Lambda and CloudWatch log group.
2. Deploy the CloudFormation stack to the source Region (us-east-1). For instructions, see Create a stack from the CloudFormation console.

This will deploy a Lambda function (book-hotel-lambda) and a CloudWatch log group (/lex/book-hotel-bot) in the us-east-1 Region.

3. Deploy the CloudFormation stack to the replica Region (us-west-2).

This will deploy a Lambda function (book-hotel-lambda) and a CloudWatch log group (/lex/book-hotel-bot) in the us-west-2 Region. The Lambda function name and CloudWatch log group name must be the same in both Regions.

4. On the Amazon Lex console in the source Region (us-east-1), navigate to the BookHotel.
5. Choose Aliases in the navigation pane, and choose the BookHotelDemoAlias_GR.
6. In the Languages section, choose English (US).
7. Select the book-hotel-lambda function and associate it with the BookHotel bot by choosing Save.
8. Navigate back to the BookHotelDemoAlias_GR alias, and in the Conversation logs section, choose Manage conversation logs.
9. Enable Text logs and select the /lex/book-hotel-bot log group, then choose Save.

Conversation text logs are now enabled for the BookHotel bot in us-east-1.

10. Switch to the replica Region (us-west-2) and navigate to the BookHotel.
11. Choose Aliases in the navigation pane, and choose the BookHotelDemoAlias_GR.

You can see that the conversation logs are already associated with the /lex/book-hotel-bot CloudWatch group the us-west-2 Region.

12. In the Languages section, choose English (US).

You can see that the book-hotel-lambda function is associated with the BookHotel alias.

Through this process, we have demonstrated how Lambda functions and CloudWatch log groups are automatically associated with the corresponding bot resources in the replica Region for the replicated bots, providing a seamless and consistent integration across both Regions.

Disabling Global Resiliency

You have the flexibility to disable Global Resiliency at any time. By disabling Global Resiliency, your source bot, along with its associated aliases and versions, will no longer be replicated across other Regions. In this section, we demonstrate the process to disable Global Resiliency.

1. On the Amazon Lex console in your source Region (us-east-1), choose Bots in the navigation pane and locate the BookHotel.
2. Under Deployment in the navigation pane, choose Global Resiliency.
3. Choose Disable Global Resiliency.

4. Enter confirm in the confirmation box and choose Delete.

This action initiates the deletion of the replicated BookHotel bot in the replica Region.

The replication status will change to Deleting, and after a few minutes, the deletion process will be complete. You will then see the Create replica option available again. If you don’t see it, try refreshing the page.

5. Check the Bot versions page of the BookHotel bot to confirm that Version 2 is still the latest version.
6. Check the Aliases page to confirm that the BookHotelDemoAlias_GR alias is still present on the source bot.

Applications referring to this alias can continue to function as normal in the source Region.

7. Switch to the replica Region (us-west-2) to confirm that the BookHotel bot has been deleted from this Region.

You can reenable Global Resiliency on the source Region (us-east-1) by going through the process described earlier in this post.

Clean up

To prevent incurring charges, complete the following steps to clean up the resources created during this demonstration:

1. Disable Global Resiliency for the bot by following the instructions detailed earlier in this post.
2. Delete the book-hotel-lambda-cw-stack CloudFormation stack from the us-west-2. For instructions, see Delete a stack on the CloudFormation console.
3. Delete the book-hotel-lambda-cw-stack CloudFormation stack from the us-east-1.
4. Delete the book-hotel-stack CloudFormation stack from the us-east-1.

Integrations with Amazon Connect

Amazon Lex Global Resiliency seamlessly complements Amazon Connect Global Resiliency, providing you with a comprehensive solution for maintaining business continuity and resilience across your conversational AI and contact center infrastructure. Amazon Connect Global Resiliency enables you to automatically maintain your instances synchronized across two Regions, making sure that all configuration resources, such as contact flows, queues, and agents, are true replicas of each other.

With the addition of Amazon Lex Global Resiliency, Amazon Connect customers gain the added benefit of automated synchronization of their Amazon Lex V2 bots associated with their contact flows. This integration provides a consistent and uninterrupted experience during failover scenarios, because your Amazon Lex interactions seamlessly transition between Regions without any disruption. By combining these complementary features, you can achieve end-to-end resilience. This minimizes the risk of downtime and makes sure your conversational AI and contact center operations remain highly available and responsive, even in the case of Regional failures or capacity constraints.

Global Resiliency APIs

Global Resiliency provides API support to create and manage replicas. These are supported in the AWS CLI and AWS SDKs. In this section, we demonstrate usage with the AWS CLI.

1. Create a bot replica in the replica Region using the CreateBotReplica.
2. Monitor the bot replication status using the DescribeBotReplica.
3. List the replicated bots using the ListBotReplicas.
4. List all the version replication statuses applicable for Global Resiliency using the ListBotVersionReplicas.

This list includes only the replicated bot versions, which were created after Global Resiliency was enabled. In the API response, a botVersionReplicationStatus of Available indicates that the bot version was replicated successfully.

5. List all the alias replication statuses applicable for Global Resiliency using the ListBotAliasReplicas.

This list includes only the replicated bot aliases, which were created after Global Resiliency was enabled. In the API response, a botAliasReplicationStatus of Available indicates that the bot alias was replicated successfully.

Conclusion

In this post, we introduced the Global Resiliency feature for Amazon Lex V2 bots. We discussed the process to enable Global Resiliency using the console and reviewed some of the new APIs released as part of this feature.

As the next step, you can explore Global Resiliency and apply the techniques discussed in this post to replicate bots and bot versions across Regions. This hands-on practice will solidify your understanding of managing and replicating Amazon Lex V2 bots in your solution architecture.

About the Authors

Priti Aryamane is a Specialty Consultant at AWS Professional Services. With over 15 years of experience in contact centers and telecommunications, Priti specializes in helping customers achieve their desired business outcomes with customer experience on AWS using Amazon Lex, Amazon Connect, and generative AI features.

Sanjeet Sanda is a Specialty Consultant at AWS Professional Services with over 20 years of experience in telecommunications, contact center technology, and customer experience. He specializes in designing and delivering customer-centric solutions with a focus on integrating and adapting existing enterprise call centers into Amazon Connect and Amazon Lex environments. Sanjeet is passionate about streamlining adoption processes by using automation wherever possible. Outside of work, Sanjeet enjoys hanging out with his family, having barbecues, and going to the beach.

Yogesh Khemka is a Senior Software Development Engineer at AWS, where he works on large language models and natural language processing. He focuses on building systems and tooling for scalable distributed deep learning training and real-time inference.

Read More

Sentence transformers are powerful deep learning models that convert sentences into high-quality, fixed-length embeddings, capturing their semantic meaning. These embeddings are useful for various natural language processing (NLP) tasks such as text classification, clustering, semantic search, and information retrieval.

In this post, we showcase how to fine-tune a sentence transformer specifically for classifying an Amazon product into its product category (such as toys or sporting goods). We showcase two different sentence transformers, paraphrase-MiniLM-L6-v2 and a proprietary Amazon large language model (LLM) called M5_ASIN_SMALL_V2.0, and compare their results. M5 LLMS are BERT-based LLMs fine-tuned on internal Amazon product catalog data using product title, bullet points, description, and more. They are currently being used for use cases such as automated product classification and similar product recommendations. Our hypothesis is that M5_ASIN_SMALL_V2.0 will perform better for the use case of Amazon product category classification due to it being fine-tuned with Amazon product data. We prove this hypothesis in the following experiment illustrated in this post.

Solution overview

In this post, we demonstrate how to fine-tune a sentence transformer with Amazon product data and how to use the resulting sentence transformer to improve classification accuracy of product categories using an XGBoost decision tree. For this demonstration, we use a public Amazon product dataset called Amazon Product Dataset 2020 from a kaggle competition. This dataset contains the following attributes and fields:

• Domain name – amazon.com
• Date range – January 1, 2020, through January 31, 2020
• File extension – CSV
• Available fields – Uniq Id, Product Name, Brand Name, Asin, Category, Upc Ean Code, List Price, Selling Price, Quantity, Model Number, About Product, Product Specification, Technical Details, Shipping Weight, Product Dimensions, Image, Variants, SKU, Product Url, Stock, Product Details, Dimensions, Color, Ingredients, Direction To Use, Is Amazon Seller, Size Quantity Variant, and Product Description
• Label field – Category

Prerequisites

Before you begin, install the following packages. You can do this in either an Amazon SageMaker notebook or your local Jupyter notebook by running the following commands:

!pip install sentencepiece --quiet
!pip install sentence_transformers --quiet
!pip install xgboost –-quiet
!pip install scikit-learn –-quiet/

Preprocess the data

The first step needed for fine-tuning a sentence transformer is to preprocess the Amazon product data for the sentence transformer to be able to consume the data and fine-tune effectively. It involves normalizing the text data, defining the product’s main category by extracting the first category from the Category field, and selecting the most important fields from the dataset that contribute to classifying the product’s main category accurately. We use the following code for preprocessing:

import pandas as pd
from sklearn.preprocessing import LabelEncoder

data = pd.read_csv('marketing_sample_for_amazon_com-ecommerce__20200101_20200131__10k_data.csv')
data.columns = data.columns.str.lower().str.replace(' ', '_')
data['main_category'] = data['category'].str.split("|").str[0]
data["all_text"] = data.apply(
    lambda r: " ".join(
        [
            str(r["product_name"]) if pd.notnull(r["product_name"]) else "",
            str(r["about_product"]) if pd.notnull(r["about_product"]) else "",
            str(r["product_specification"]) if pd.notnull(r["product_specification"]) else "",
            str(r["technical_details"]) if pd.notnull(r["technical_details"]) else ""
        ]
    ),
    axis=1
)
label_encoder = LabelEncoder()
labels_transform = label_encoder.fit_transform(data['main_category'])
data['label']=labels_transform
data[['all_text','label']]

The following screenshot shows an example of what our dataset looks like after it has been preprocessed.

Fine-tune the sentence transformer paraphrase-MiniLM-L6-v2

The first sentence transformer we fine-tune is called paraphrase-MiniLM-L6-v2. It uses the popular BERT model as its underlying architecture to transform product description text into a 384-dimensional dense vector embedding that will be consumed by our XGBoost classifier for product category classification. We use the following code to fine-tune paraphrase-MiniLM-L6-v2 using the preprocessed Amazon product data:

from sentence_transformers import SentenceTransformer
model_name='paraphrase-MiniLM-L6-v2'
model = SentenceTransformer(model_name)

The first step is to define a classification head that represents the 24 product categories that an Amazon product can be classified into. This classification head will be used to train the sentence transformer specifically to be more effective at transforming product descriptions according to the 24 product categories. The idea is that all product descriptions that are within the same category should be transformed into a vector embedding that is closer in distance compared to product descriptions that belong in different categories.

 The following code is for fine-tuning sentence transformer 1:

import torch.nn as nn

# Define classification head
class ClassificationHead(nn.Module):
    def __init__(self, embedding_dim, num_classes):
        super(ClassificationHead, self).__init__()
        self.linear = nn.Linear(embedding_dim, num_classes)

    def forward(self, features):
        x = features['sentence_embedding']
        x = self.linear(x)
        return x

# Define the number of classes for a classification task.
num_classes = 24
print('class number:', num_classes)
classification_head = ClassificationHead(model.get_sentence_embedding_dimension(), num_classes)

# Combine SentenceTransformer model and classification head."
class SentenceTransformerWithHead(nn.Module):
    def __init__(self, transformer, head):
        super(SentenceTransformerWithHead, self).__init__()
        self.transformer = transformer
        self.head = head

    def forward(self, input):
        features = self.transformer(input)
        logits = self.head(features)
        return logits

model_with_head = SentenceTransformerWithHead(model, classification_head)

We then set the fine-tuning parameters. For this post, we train on five epochs, optimize for cross-entropy loss, and use the AdamW optimization method. We chose epoch 5 because, after testing various epoch values, we observed that the loss minimized at epoch 5. This made it the optimal number of training iterations for achieving the best classification results.

The following code is for fine-tuning sentence transformer 2:

import os
os.environ["TORCH_USE_CUDA_DSA"] = "1"
os.environ["CUDA_LAUNCH_BLOCKING"] = "1"

from sentence_transformers import SentenceTransformer, InputExample, LoggingHandler
import torch
from torch.utils.data import DataLoader
from transformers import AdamW, get_linear_schedule_with_warmup

train_sentences = data['all_text']
train_labels = data['label']
# training parameters
num_epochs = 5
batch_size = 2
learning_rate = 2e-5

# Convert the dataset to PyTorch tensors.
train_examples = [InputExample(texts=[s], label=l) for s, l in zip(train_sentences, train_labels)]

# Customize collate_fn to convert InputExample objects into tensors.
def collate_fn(batch):
    texts = [example.texts[0] for example in batch]
    labels = torch.tensor([example.label for example in batch])
    return texts, labels

train_dataloader = DataLoader(train_examples, shuffle=True, batch_size=batch_size, collate_fn=collate_fn)

# Define the loss function, optimizer, and learning rate scheduler.
criterion = nn.CrossEntropyLoss()
optimizer = AdamW(model_with_head.parameters(), lr=learning_rate)
total_steps = len(train_dataloader) * num_epochs
scheduler = get_linear_schedule_with_warmup(optimizer, num_warmup_steps=0, num_training_steps=total_steps)

# Training loop
loss_list=[]
for epoch in range(num_epochs):
    model_with_head.train()
    for step, (texts, labels) in enumerate(train_dataloader):
        labels = labels.to(model.device)
        optimizer.zero_grad()

        # Encode text and pass through classification head.
        inputs = model.tokenize(texts)
        input_ids = inputs['input_ids'].to(model.device)
        input_attention_mask = inputs['attention_mask'].to(model.device)
        inputs_final = {'input_ids': input_ids, 'attention_mask': input_attention_mask}
        
        # move model_with_head to the same device
        model_with_head = model_with_head.to(model.device)
        logits = model_with_head(inputs_final)
        
        loss = criterion(logits, labels)
        loss.backward()
        optimizer.step()
        scheduler.step()
        if step % 100 == 0:
            print(f"Epoch {epoch}, Step {step}, Loss: {loss.item()}")

    print(f'Epoch {epoch+1}/{num_epochs}, Loss: {loss.item()}')
    model_save_path = f'./intermediate-output/epoch-{epoch}'
    model.save(model_save_path)
    loss_list.append(loss.item())
# Save the final model
model_final_save_path='st_ft_epoch_5'
model.save(model_final_save_path)

To observe whether our resulting fine-tuned sentence transformer improves our product category classification accuracy, we use it as our text embedder in the XGBoost classifier in the next step.

XGBoost classification

XGBoost (Extreme Gradient Boosting) classification is a machine learning technique used for classification tasks. It’s an implementation of the gradient boosting framework designed to be efficient, flexible, and portable. For this post, we have XGBoost consume the product description text embedding output of our sentence transformers and observe product category classification accuracy. We use the following code to use the standard paraphrase-MiniLM-L6-v2 sentence transformer before it was fine-tuned to classify Amazon products to their respective categories:

from sklearn.model_selection import train_test_split
import xgboost as xgb
from sklearn.metrics import accuracy_score

model = SentenceTransformer('paraphrase-MiniLM-L6-v2')  
data['text_embedding'] = data['all_text'].apply(lambda x: model.encode(str(x)))
text_embeddings = pd.DataFrame(data['text_embedding'].tolist(), index=data.index, dtype=float)

# Convert numeric columns stored as strings to floats
numeric_columns = ['selling_price', 'shipping_weight', 'product_dimensions']  # Add more columns as needed
for col in numeric_columns:
    data[col] = pd.to_numeric(data[col], errors='coerce')

# Convert categorical columns to category type
categorical_columns = ['model_number', 'is_amazon_seller']  # Add more columns as needed
for col in categorical_columns:
    data[col] = data[col].astype('category')
    
X_0 = data[['selling_price','model_number','is_amazon_seller']]
X = pd.concat([X_0, text_embeddings], axis=1)
label_encoder = LabelEncoder()
data['main_category_encoded'] = label_encoder.fit_transform(data['main_category'])
y = data['main_category_encoded']
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)

# Re-encode the labels to ensure they are consecutive integers starting from 0
unique_labels = sorted(set(y_train) | set(y_test))
label_mapping = {label: idx for idx, label in enumerate(unique_labels)}

y_train = y_train.map(label_mapping)
y_test = y_test.map(label_mapping)

# Enable categorical support for XGBoost
dtrain = xgb.DMatrix(X_train, label=y_train, enable_categorical=True)
dtest = xgb.DMatrix(X_test, label=y_test, enable_categorical=True)

param = {
    'max_depth': 6,
    'eta': 0.3,
    'objective': 'multi:softmax',
    'num_class': len(label_mapping),
    'eval_metric': 'mlogloss'
}

num_round = 100
bst = xgb.train(param, dtrain, num_round)

# Evaluate the model
y_pred = bst.predict(dtest)
accuracy = accuracy_score(y_test, y_pred)
print(f'Accuracy: {accuracy:.2f}')

Accuracy: 0.78

We observe a 78% accuracy using the stock paraphrase-MiniLM-L6-v2 sentence transformer. To observe the results of the fine-tuned paraphrase-MiniLM-L6-v2 sentence transformer, we need to update the beginning of the code as follows. All other code remains the same.

model = SentenceTransformer('st_ft_epoch_5')  
data['text_embedding_miniLM_ft10'] = data['all_text'].apply(lambda x: model.encode(str(x)))
text_embeddings = pd.DataFrame(data['text_embedding_finetuned'].tolist(), index=data.index, dtype=float)
X_pa_finetuned = pd.concat([X_0, text_embeddings], axis=1)
X_train, X_test, y_train, y_test = train_test_split(X_pa_finetuned, y, test_size=0.2, random_state=42)

# Re-encode the labels to ensure they are consecutive integers starting from 0
unique_labels = sorted(set(y_train) | set(y_test))
label_mapping = {label: idx for idx, label in enumerate(unique_labels)}

y_train = y_train.map(label_mapping)
y_test = y_test.map(label_mapping)

# Build and train the XGBoost model
# Enable categorical support for XGBoost
dtrain = xgb.DMatrix(X_train, label=y_train, enable_categorical=True)
dtest = xgb.DMatrix(X_test, label=y_test, enable_categorical=True)

param = {
    'max_depth': 6,
    'eta': 0.3,
    'objective': 'multi:softmax',
    'num_class': len(label_mapping),
    'eval_metric': 'mlogloss'
}

num_round = 100
bst = xgb.train(param, dtrain, num_round)

y_pred = bst.predict(dtest)
accuracy = accuracy_score(y_test, y_pred)
print(f'Accuracy: {accuracy:.2f}')

# Optionally, convert the predicted labels back to the original category labels
inverse_label_mapping = {idx: label for label, idx in label_mapping.items()}
y_pred_labels = pd.Series(y_pred).map(inverse_label_mapping)

Accuracy: 0.94

With the fine-tuned paraphrase-MiniLM-L6-v2 sentence transformer, we observe a 94% accuracy, a 16% increase from the baseline of 78% accuracy. From this observation, we conclude that fine-tuning paraphrase-MiniLM-L6-v2 is effective for classifying Amazon product data into product categories.

Fine-tune the sentence transformer M5_ASIN_SMALL_V20

Now we create a sentence transformer from a BERT-based model called M5_ASIN_SMALL_V2.0. It’s a 40-million-parameter BERT-based model trained at M5, an internal team at Amazon specializing in fine-tuning LLMs using Amazon product data. It was distilled from a larger teacher model (approximately 5 billion parameters), which was pre-trained on a large amount of unlabeled ASIN data and pre-fine-tuned on a set of Amazon supervised learning tasks (multi-task pre-fine-tuning). It is a multi-task, multi-lingual, multi-locale, and multi-modal BERT-based encoder-only model trained on text and structured data input. Its neural network architectural details are as follows:

Model backbone:
 Hidden size: 384
 Number of hidden layers: 24
 Number of attention heads: 16
 Intermediate size: 1536
 Vocabulary size: 256,035
Number of backbone parameters: 42,587,904
Number of word embedding parameters (bert.embedding.*): 98,517,504
Total number of parameters: 141,259,023

Because M5_ASIN_SMALL_V20 was pre-trained on Amazon product data specifically, we hypothesize that building a sentence transformer from it will increase the accuracy of product category classification. We complete the following steps to build a sentence transformer from M5_ASIN_SMALL_V20, fine-tune it, and input it into an XGBoost classifier to observe accuracy impact:

1. Load a pre-trained M5 model that you want to use as the base encoder.
2. Use the M5 model within the SentenceTransformer framework to create a sentence transformer.
3. Add a pooling layer to create fixed-size sentence embeddings from the variable-length output of the BERT model.
4. Combine the M5 model and pooling layer into a single model.
5. Fine-tune the model on a relevant dataset.

See the following code for Steps 1–3:

from sentence_transformers import models 
from transformers import AutoTokenizer

# Step 1: Load Pre-trained M5 Model
model_path = 'M5_ASIN_SMALL_V20'  # or your custom model path
transformer_model = models.Transformer(model_path)
tokenizer = AutoTokenizer.from_pretrained(model_path)

# Step 2: Define Pooling Layer
pooling_model = models.Pooling(transformer_model.get_word_embedding_dimension(),
                               pooling_mode_mean_tokens=True)

# Step 3: Create SentenceTransformer Model
model_mean_m5_base = SentenceTransformer(modules=[transformer_model, pooling_model])

The rest of the code remains the same as fine-tuning for the paraphrase-MiniLM-L6-v2 sentence transformer, except that we use the fine-tuned M5 sentence transformer instead to create embeddings for the texts in the dataset:

loaded_model = SentenceTransformer('m5_ft_epoch_5_mean')
data['text_embedding_m5'] = data['all_text'].apply(lambda x: loaded_model.encode(str(x)))

Result

We observe similar results to paraphrase-MiniLM-L6-v2 when looking at accuracy before fine-tuning, observing a 78% accuracy for M5_ASIN_SMALL_V20. However, we observe that the fine-tuned M5_ASIN_SMALL_V20 sentence transformer performs better than the fine-tuned paraphrase-MiniLM-L6-v2. Its accuracy is 98%, compared to 94% for the fine-tuned paraphrase-MiniLM-L6-v2. We fine-tuned the sentence transformers for 5 epochs, because experiments showed this was the optimal number to minimize loss. The following graph summarizes our observations of accuracy improvement with fine-tuning for 5 epochs in a single comparison chart.

Clean up

We recommend using GPUs to fine-tune the sentence transformers, for example, ml.g5.4xlarge or ml.g4dn.16xlarge. Be sure to clean up resources to avoid incurring additional costs.

If you’re using a SageMaker notebook instance, refer to Clean up Amazon SageMaker notebook instance resources. If you’re using Amazon SageMaker Studio, refer to Delete or stop your Studio running instances, applications, and spaces.

Conclusion

In this post, we explored sentence transformers and how to use them effectively for text classification tasks. We dived deep into the sentence transformer paraphrase-MiniLM-L6-v2, demonstrated how to use a BERT-based model like M5_ASIN_SMALL_V20 to create a sentence transformer, showed how to fine-tune sentence transformers, and showed the accuracy effects of fine-tuning sentence transformers.

Fine-tuning sentence transformers has proven to be highly effective for classifying product descriptions into categories, significantly enhancing prediction accuracy. As a next step, we encourage you to explore different sentence transformers from Hugging Face.

Lastly, if you want to explore M5, note that it is proprietary to Amazon and you can only access it as an Amazon partner or customer as of the time of this publication. Connect with your Amazon point of contact if you’re an Amazon partner or customer wanting to use M5, and they will guide you through M5’s offerings and how it can be used for your use case.

About the Authors

Kara Yang is a Data Scientist at AWS Professional Services in the San Francisco Bay Area, with extensive experience in AI/ML. She specializes in leveraging cloud computing, machine learning, and Generative AI to help customers address complex business challenges across various industries. Kara is passionate about innovation and continuous learning.

Farshad Harirchi is a Principal Data Scientist at AWS Professional Services. He helps customers across industries, from retail to industrial and financial services, with the design and development of generative AI and machine learning solutions. Farshad brings extensive experience in the entire machine learning and MLOps stack. Outside of work, he enjoys traveling, playing outdoor sports, and exploring board games.

James Poquiz is a Data Scientist with AWS Professional Services based in Orange County, California. He has a BS in Computer Science from the University of California, Irvine and has several years of experience working in the data domain having played many different roles. Today he works on implementing and deploying scalable ML solutions to achieve business outcomes for AWS clients.

Read More

Recently, we’ve been witnessing the rapid development and evolution of generative AI applications, with observability and evaluation emerging as critical aspects for developers, data scientists, and stakeholders. Observability refers to the ability to understand the internal state and behavior of a system by analyzing its outputs, logs, and metrics. Evaluation, on the other hand, involves assessing the quality and relevance of the generated outputs, enabling continual improvement.

Comprehensive observability and evaluation are essential for troubleshooting, identifying bottlenecks, optimizing applications, and providing relevant, high-quality responses. Observability empowers you to proactively monitor and analyze your generative AI applications, and evaluation helps you collect feedback, refine models, and enhance output quality.

In the context of Amazon Bedrock, observability and evaluation become even more crucial. Amazon Bedrock is a fully managed service that offers a choice of high-performing foundation models (FMs) from leading AI companies such as AI21 Labs, Anthropic, Cohere, Meta, Stability AI, and Amazon through a single API, along with a broad set of capabilities you need to build generative AI applications with security, privacy, and responsible AI. As the complexity and scale of these applications grow, providing comprehensive observability and robust evaluation mechanisms are essential for maintaining high performance, quality, and user satisfaction.

We have built a custom observability solution that Amazon Bedrock users can quickly implement using just a few key building blocks and existing logs using FMs, Amazon Bedrock Knowledge Bases, Amazon Bedrock Guardrails, and Amazon Bedrock Agents. This solution uses decorators in your application code to capture and log metadata such as input prompts, output results, run time, and custom metadata, offering enhanced security, ease of use, flexibility, and integration with native AWS services.

Notably, the solution supports comprehensive Retrieval Augmented Generation (RAG) evaluation so you can assess the quality and relevance of generated responses, identify areas for improvement, and refine the knowledge base or model accordingly.

In this post, we set up the custom solution for observability and evaluation of Amazon Bedrock applications. Through code examples and step-by-step guidance, we demonstrate how you can seamlessly integrate this solution into your Amazon Bedrock application, unlocking a new level of visibility, control, and continual improvement for your generative AI applications.

By the end of this post, you will:

1. Understand the importance of observability and evaluation in generative AI applications
2. Learn about the key features and benefits of this solution
3. Gain hands-on experience in implementing the solution through step-by-step demonstrations
4. Explore best practices for integrating observability and evaluation into your Amazon Bedrock workflows

Prerequisites

To implement the observability solution discussed in this post, you need the following prerequisites:

• An active Amazon Web Services (AWS) account and AWS Identity and Access Management (IAM) role with Amazon Bedrock access
• Access to the FMs you plan to use
• Basic understanding of decorators in your preferred programming language (Python or Node.js)
• A clone of the amazon-bedrock-samples GitHub repository
• Basic familiarity with AWS services such as Amazon Data Firehose, Amazon Athena, and AWS Glue crawlers (optional, depending on the specific components used in the solution)

Solution overview

The observability solution for Amazon Bedrock empowers users to track and analyze interactions with FMs, knowledge bases, guardrails, and agents using decorators in their source code. Key highlights of the solution include:

• Decorator – Decorators are applied to functions invoking Amazon Bedrock APIs, capturing input prompt, output results, custom metadata, custom metrics, and latency related metrics.
• Flexible logging –You can use this solution to store logs either locally or in Amazon Simple Storage Service (Amazon S3) using Amazon Data Firehose, enabling integration with existing monitoring infrastructure. Additionally, you can choose what gets logged.
• Dynamic data partitioning – The solution enables dynamic partitioning of observability data based on different workflows or components of your application, such as prompt preparation, data preprocessing, feedback collection, and inference. This feature allows you to separate data into logical partitions, making it easier to analyze and process data later.
• Security – The solution uses AWS services and adheres to AWS Cloud Security best practices so your data remains within your AWS account.
• Cost optimization – This solution uses serverless technologies, making it cost-effective for the observability infrastructure. However, some components may incur additional usage-based costs.
• Multiple programming language support – The GitHub repository provides the observability solution in both Python and Node.js versions, catering to different programming preferences.

Here’s a high-level overview of the observability solution architecture:

The following steps explain how the solution works:

1. Application code using Amazon Bedrock is decorated with @bedrock_logs.watch to save the log
2. Logged data streams through Amazon Data Firehose
3. AWS Lambda transforms the data and applies dynamic partitioning based on call_type variable
4. Amazon S3 stores the data securely
5. Optional components for advanced analytics
6. AWS Glue creates tables from S3 data
7. Amazon Athena enables data querying
8. Visualize logs and insights in your favorite dashboard tool

This architecture provides comprehensive logging, efficient data processing, and powerful analytics capabilities for your Amazon Bedrock applications.

Getting started

To help you get started with the observability solution, we have provided example notebooks in the attached GitHub repository, covering knowledge bases, evaluation, and agents for Amazon Bedrock. These notebooks demonstrate how to integrate the solution into your Amazon Bedrock application and showcase various use cases and features including feedback collected from users or quality assurance (QA) teams.

The repository contains well-documented notebooks that cover topics such as:

• Setting up the observability infrastructure
• Integrating the decorator pattern into your application code
• Logging model inputs, outputs, and custom metadata
• Collecting and analyzing feedback data
• Evaluating model responses and knowledge base performance
• Example visualization for observability data using AWS services

To get started with the example notebooks, follow these steps:

1. Clone the GitHub repository

   git clone https://github.com/aws-samples/amazon-bedrock-samples.git

2. Navigate to the observability solution directory

   cd amazon-bedrock-samples/evaluation-observe/Custom-Observability-Solution

3. Follow the instructions in the README file to set up the required AWS resources and configure the solution
4. Open the provided Jupyter notebooks and follow along with the examples and demonstrations

These notebooks provide a hands-on learning experience and serve as a starting point for integrating our solution into your generative AI applications. Feel free to explore, modify, and adapt the code examples to suit your specific requirements.

Key features

The solution offers a range of powerful features to streamline observability and evaluation for your generative AI applications on Amazon Bedrock:

• Decorator-based implementation – Use decorators to seamlessly integrate observability logging into your application functions, capturing inputs, outputs, and metadata without modifying the core logic
• Selective logging – Choose what to log by selectively capturing function inputs, outputs, or excluding sensitive information or large data structures that might not be relevant for observability
• Logical data partitioning – Create logical partitions in the observability data based on different workflows or application components, enabling easier analysis and processing of specific data subsets
• Human-in-the-loop evaluation – Collect and associate human feedback with specific model responses or sessions, facilitating comprehensive evaluation and continual improvement of your application’s performance and output quality
• Multi-component support – Support observability and evaluation for various Amazon Bedrock components, including InvokeModel, batch inference, knowledge bases, agents, and guardrails, providing a unified solution for your generative AI applications
• Comprehensive evaluation – Evaluate the quality and relevance of generated responses, including RAG evaluation for knowledge base applications, using the open source RAGAS library to compute evaluation metrics

This concise list highlights the key features you can use to gain insights, optimize performance, and drive continual improvement for your generative AI applications on Amazon Bedrock. For a detailed breakdown of the features and implementation specifics, refer to the comprehensive documentation in the GitHub repository.

Implementation and best practices

The solution is designed to be modular and flexible so you can customize it according to your specific requirements. Although the implementation is straightforward, following best practices is crucial for the scalability, security, and maintainability of your observability infrastructure.

Solution deployment

This solution includes an AWS CloudFormation template that streamlines the deployment of required AWS resources, providing consistent and repeatable deployments across environments. The CloudFormation template provisions resources such as Amazon Data Firehose delivery streams, AWS Lambda functions, Amazon S3 buckets, and AWS Glue crawlers and databases.

Decorator pattern

The solution uses the decorator pattern to integrate observability logging into your application functions seamlessly. The @bedrock_logs.watch decorator wraps your functions, automatically logging inputs, outputs, and metadata to Amazon Kinesis Firehose. Here’s an example of how to use the decorator:

# import observability
from observability import BedrockLogs

# instantiate BedrockLogs in Firehose mode
bedrock_logs = BedrockLogs(delivery_stream_name='your-firehose-delivery-stream', feedback_variables=True)

# decorate your function
@bedrock_logs.watch(capture_input=True, capture_output=True, call_type='<your-custom-dataset-name>')
def your_function(arg1, arg2):
    # Your function code here along with any custom metric of your choosing
    return output

Human-in-the-loop evaluation

The solution supports human-in-the-loop evaluation so you can incorporate human feedback into the performance evaluation of your generative AI application. You can involve end users, experts, or QA teams in the evaluation process, providing insights to enhance output quality and relevance. Here’s an example of how you can implement human-in-the-loop evaluation:

@bedrock_logs.watch(call_type='Retrieve-and-Generate-with-KB')
def main(input_arguments):
    # Your code to interact with Amazon Bedrock Knowledge Base or Agent
    return response, custom_metric, etc.

@bedrock_logs.watch(call_type='observation-feedback')
def observation_level_feedback(feedback):
    pass

# Invoke main function with user input and get run_id and observation_id
tuple_of_function_outputs, run_id, observation_id = main(input_arguments)

# Collect human feedback on model response in your application
user_feedback = 'thumbs-up'

observation_feedback_from_front_end = {
    'user_id': 'User-1',
    'f_run_id': run_id,
    'f_observation_id': observation_id,
    'actual_feedback': user_feedback
}

# Log the human-in-loop feedback using observation_level_feedback function
observation_level_feedback(observation_feedback_from_front_end)

By using the run_id and observation_id generated, you can associate human feedback with specific model responses or sessions. This feedback can then be analyzed and used to refine the knowledge base, fine-tune models, or identify areas for improvement.

Best practices

It’s recommended to follow these best practices:

• Plan call types in advance – Determine the logical partitions (call_type) for your observability data based on different workflows or application components. This enables easier analysis and processing of specific data subsets.
• Use feedback variables – Configure feedback_variables=True when initializing BedrockLogs to generate run_id and observation_id. These IDs can be used to join logically partitioned datasets, associating feedback data with corresponding model responses.
• Extend for general steps – Although the solution is designed for Amazon Bedrock, you can use the decorator pattern to log observability data for general steps such as prompt preparation, postprocessing, or other custom workflows.
• Log custom metrics – If you need to calculate custom metrics such as latency, context relevance, faithfulness, or any other metric, you can pass these values in the response of your decorated function, and the solution will log them alongside the observability data.
• Selective logging – Use the capture_input and capture_output parameters to selectively log function inputs or outputs or exclude sensitive information or large data structures that might not be relevant for observability.
• Comprehensive evaluation – Evaluate the quality and relevance of generated responses, including RAG evaluation for knowledge base applications, using the KnowledgeBasesEvaluations

By following these best practices and using the features of the solution, you can set up comprehensive observability and evaluation for your generative AI applications to gain valuable insights, identify areas for improvement, and enhance the overall user experience.

In the next post in this three-part series, we dive deeper into observability and evaluation for RAG and agent-based generative AI applications, providing in-depth insights and guidance.

Clean up

To avoid incurring costs and maintain a clean AWS account, you can remove the associated resources by deleting the AWS CloudFormation stack you created for this walkthrough. You can follow the steps provided in the Deleting a stack on the AWS CloudFormation console documentation to delete the resources created for this solution.

Conclusion and next steps

This comprehensive solution empowers you to seamlessly integrate comprehensive observability into your generative AI applications in Amazon Bedrock. Key benefits include streamlined integration, selective logging, custom metadata tracking, and comprehensive evaluation capabilities, including RAG evaluation. Use AWS services such as Athena to analyze observability data, drive continual improvement, and connect with your favorite dashboard tool to visualize the data.

This post focused is on Amazon Bedrock, but it can be extended to broader machine learning operations (MLOps) workflows or integrated with other AWS services such as AWS Lambda or Amazon SageMaker. We encourage you to explore this solution and integrate it into your workflows. Access the source code and documentation in our GitHub repository  and start your integration journey. Embrace the power of observability and unlock new heights for your generative AI applications.

About the authors

Ishan Singh is a Generative AI Data Scientist at Amazon Web Services, where he helps customers build innovative and responsible generative AI solutions and products. With a strong background in AI/ML, Ishan specializes in building Generative AI solutions that drive business value. Outside of work, he enjoys playing volleyball, exploring local bike trails, and spending time with his wife and dog, Beau.

Chris Pecora is a Generative AI Data Scientist at Amazon Web Services. He is passionate about building innovative products and solutions while also focused on customer-obsessed science. When not running experiments and keeping up with the latest developments in generative AI, he loves spending time with his kids.

Yanyan Zhang is a Senior Generative AI Data Scientist at Amazon Web Services, where she has been working on cutting-edge AI/ML technologies as a Generative AI Specialist, helping customers use generative AI to achieve their desired outcomes. Yanyan graduated from Texas A&M University with a PhD in Electrical Engineering. Outside of work, she loves traveling, working out, and exploring new things.

Mani Khanuja is a Tech Lead – Generative AI Specialists, author of the book Applied Machine Learning and High Performance Computing on AWS, and a member of the Board of Directors for Women in Manufacturing Education Foundation Board. She leads machine learning projects in various domains such as computer vision, natural language processing, and generative AI. She speaks at internal and external conferences such AWS re:Invent, Women in Manufacturing West, YouTube webinars, and GHC 23. In her free time, she likes to go for long runs along the beach.

Read More

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

Batch inference in Amazon Bedrock efficiently processes large volumes of data using foundation models (FMs) when real-time results aren’t necessary. It’s ideal for workloads that aren’t latency sensitive, such as obtaining embeddings, entity extraction, FM-as-judge evaluations, and text categorization and summarization for business reporting tasks. A key advantage is its cost-effectiveness, with batch inference workloads charged at a 50% discount compared to On-Demand pricing. Refer to Supported Regions and models for batch inference for current supporting AWS Regions and models.

Although batch inference offers numerous benefits, it’s limited to 10 batch inference jobs submitted per model per Region. To address this consideration and enhance your use of batch inference, we’ve developed a scalable solution using AWS Lambda and Amazon DynamoDB. This post guides you through implementing a queue management system that automatically monitors available job slots and submits new jobs as slots become available.

We walk you through our solution, detailing the core logic of the Lambda functions. By the end, you’ll understand how to implement this solution so you can maximize the efficiency of your batch inference workflows on Amazon Bedrock. For instructions on how to start your Amazon Bedrock batch inference job, refer to Enhance call center efficiency using batch inference for transcript summarization with Amazon Bedrock.

The power of batch inference

Organizations can use batch inference to process large volumes of data asynchronously, making it ideal for scenarios where real-time results are not critical. This capability is particularly useful for tasks such as asynchronous embedding generation, large-scale text classification, and bulk content analysis. For instance, businesses can use batch inference to generate embeddings for vast document collections, classify extensive datasets, or analyze substantial amounts of user-generated content efficiently.

One of the key advantages of batch inference is its cost-effectiveness. Amazon Bedrock offers select FMs for batch inference at 50% of the On-Demand inference price. Organizations can process large datasets more economically because of this significant cost reduction, making it an attractive option for businesses looking to optimize their generative AI processing expenses while maintaining the ability to handle substantial data volumes.

Solution overview

The solution presented in this post uses batch inference in Amazon Bedrock to process many requests efficiently using the following solution architecture.

This architecture workflow includes the following steps:

1. A user uploads files to be processed to an Amazon Simple Storage Service (Amazon S3) bucket br-batch-inference-{Account_Id}-{AWS-Region} in the to-process folder. Amazon S3 invokes the {stack_name}-create-batch-queue-{AWS-Region} Lambda function.
2. The invoked Lambda function creates new job entries in a DynamoDB table with the status as Pending. The DynamoDB table is crucial for tracking and managing the batch inference jobs throughout their lifecycle. It stores information such as job ID, status, creation time, and other metadata.
3. The Amazon EventBridge rule scheduled to run every 15 minutes invokes the {stack_name}-process-batch-jobs-{AWS-Region} Lambda function.
4. The {stack_name}-process-batch-jobs-{AWS-Region} Lambda function performs several key tasks:
   • Scans the DynamoDB table for jobs in InProgress, Submitted, Validation and Scheduled status
   • Updates job status in DynamoDB based on the latest information from Amazon Bedrock
   • Calculates available job slots and submits new jobs from the Pending queue if slots are available
   • Handles error scenarios by updating job status to Failed and logging error details for troubleshooting
5. The Lambda function makes the GetModelInvocationJob API call to get the latest status of the batch inference jobs from Amazon Bedrock
6. The Lambda function then updates the status of the jobs in DynamoDB using the UpdateItem API call, making sure that the table always reflects the most current state of each job
7. The Lambda function calculates the number of available slots before the Service Quota Limit for batch inference jobs is reached. Based on this, it queries for jobs in the Pending state that can be submitted
8. If there is a slot available, the Lambda function will make CreateModelInvocationJob API calls to create new batch inference jobs for the pending jobs
9. It updates the DynamoDB table with the status of the batch inference jobs created in the previous step
10. After one batch job is complete, its output files will be available in the S3 bucket br-batch-inference-{Account_Id}-{AWS-Region} processed folder

Prerequisites

To perform the solution, you need the following prerequisites:

• An active AWS account.
• An AWS Region from the list of batch inference supported Regions for Amazon Bedrock.
• Access to your selected models hosted on Amazon Bedrock. Make sure the selected model has been enabled in Amazon Bedrock.
• If you plan to use your own AWS Identity and Access Management (IAM) role for batch inference, create it with a trust policy and Amazon S3 access (read access to the folder containing input data and write access to the folder storing output data).

Deployment guide

To deploy the pipeline, complete the following steps:

1. Choose the Launch Stack button:
   
2. Choose Next, as shown in the following screenshot
   
3. Specify the pipeline details with the options fitting your use case:
   • Stack name (Required) – The name you specified for this AWS CloudFormation. The name must be unique in the region in which you’re creating it.
   • ModelId (Required) – Provide the model ID that you need your batch job to run with.
   • RoleArn (Optional) – By default, the CloudFormation stack will deploy a new IAM role with the required permissions. If you have a role you want to use instead of creating a new role, provide the IAM role Amazon Resource Name (ARN) that has sufficient permission to create a batch inference job in Amazon Bedrock and read/write in the created S3 bucket br-batch-inference-{Account_Id}-{AWS-Region}. Follow the instructions in the prerequisites section to create this role.

4. In the Amazon Configure stack options section, add optional tags, permissions, and other advanced settings if needed. Or you can just leave it blank and choose Next, as shown in the following screenshot.
   
5. Review the stack details and select I acknowledge that AWS CloudFormation might create AWS IAM resources, as shown in the following screenshot.
   
6. Choose Submit. This initiates the pipeline deployment in your AWS account.
7. After the stack is deployed successfully, you can start using the pipeline. First, create a /to-process folder under the created Amazon S3 location for input. A .jsonl uploaded to this folder will have a batch job created with the selected model. The following is a screenshot of the DynamoDB table where you can track the job status and other types of metadata related to the job.
   
8. After your first batch job from the pipeline is complete, the pipeline will create a /processed folder under the same bucket, as shown in the following screenshot. Outputs from the batch jobs created by this pipeline will be stored in this folder.
   
9. To start using this pipeline, upload the .jsonl files you’ve prepared for batch inference in Amazon Bedrock

You’re done! You’ve successfully deployed your pipeline and you can check the batch job status in the Amazon Bedrock console. If you want to have more insights about each .jsonl file’s status, navigate to the created DynamoDB table {StackName}-DynamoDBTable-{UniqueString} and check the status there. You may need to wait up to 15 minutes to observe the batch jobs created because EventBridge is scheduled to scan DynamoDB every 15 minutes.

Clean up

If you no longer need this automated pipeline, follow these steps to delete the resources it created to avoid additional cost:

1. On the Amazon S3 console, manually delete the contents inside buckets. Make sure the bucket is empty before moving to step 2.
2. On the AWS CloudFormation console, choose Stacks in the navigation pane.
3. Select the created stack and choose Delete, as shown in the following screenshot.
   

This automatically deletes the deployed stack.

Conclusion

In this post, we’ve introduced a scalable and efficient solution for automating batch inference jobs in Amazon Bedrock. By using AWS Lambda, Amazon DynamoDB, and Amazon EventBridge, we’ve addressed key challenges in managing large-scale batch processing workflows.

This solution offers several significant benefits:

1. Automated queue management – Maximizes throughput by dynamically managing job slots and submissions
2. Cost optimization – Uses the 50% discount on batch inference pricing for economical large-scale processing

This automated pipeline significantly enhances your ability to process large amounts of data using batch inference for Amazon Bedrock. Whether you’re generating embeddings, classifying text, or analyzing content in bulk, this solution offers a scalable, efficient, and cost-effective approach to batch inference.

As you implement this solution, remember to regularly review and optimize your configuration based on your specific workload patterns and requirements. With this automated pipeline and the power of Amazon Bedrock, you’re well-equipped to tackle large-scale AI inference tasks efficiently and effectively. We encourage you to try it out and share your feedback to help us continually improve this solution.

For additional resources, refer to the following:

• User guide – Process multiple prompts with batch inference
• Code sample – Sample for building your batch inference job
• Blog post – Enhance call center efficiency using batch inference for transcript summarization with Amazon Bedrock

About the authors

Yanyan Zhang is a Senior Generative AI Data Scientist at Amazon Web Services, where she has been working on cutting-edge AI/ML technologies as a Generative AI Specialist, helping customers use generative AI to achieve their desired outcomes. Yanyan graduated from Texas A&M University with a PhD in Electrical Engineering. Outside of work, she loves traveling, working out, and exploring new things.

Ishan Singh is a Generative AI Data Scientist at Amazon Web Services, where he helps customers build innovative and responsible generative AI solutions and products. With a strong background in AI/ML, Ishan specializes in building Generative AI solutions that drive business value. Outside of work, he enjoys playing volleyball, exploring local bike trails, and spending time with his wife and dog, Beau.

Neeraj Lamba is a Cloud Infrastructure Architect with Amazon Web Services (AWS) Worldwide Public Sector Professional Services. He helps customers transform their business by helping design their cloud solutions and offering technical guidance. Outside of work, he likes to travel, play Tennis and experimenting with new technologies.

Read More

Professionals in a wide variety of industries have adopted digital video conferencing tools as part of their regular meetings with suppliers, colleagues, and customers. These meetings often involve exchanging information and discussing actions that one or more parties must take after the session. The traditional way to make sure information and actions aren’t forgotten is to take notes during the session; a manual and tedious process that can be error-prone, particularly in a high-activity or high-pressure scenario. Furthermore, these notes are usually personal and not stored in a central location, which is a lost opportunity for businesses to learn what does and doesn’t work, as well as how to improve their sales, purchasing, and communication processes.

This post presents a solution where you can upload a recording of your meeting (a feature available in most modern digital communication services such as Amazon Chime) to a centralized video insights and summarization engine. This engine uses artificial intelligence (AI) and machine learning (ML) services and generative AI on AWS to extract transcripts, produce a summary, and provide a sentiment for the call. The solution notes the logged actions per individual and provides suggested actions for the uploader. All of this data is centralized and can be used to improve metrics in scenarios such as sales or call centers. Many commercial generative AI solutions available are expensive and require user-based licenses. In contrast, our solution is an open-source project powered by Amazon Bedrock, offering a cost-effective alternative without those limitations.

This solution can help your organizations’ sales, sales engineering, and support functions become more efficient and customer-focused by reducing the need to take notes during customer calls.

Use case overview

The organization in this scenario has noticed that during customer calls, some actions often get skipped due to the complexity of the discussions, and that there might be potential to centralize customer data to better understand how to improve customer interactions in the long run. The organization already records sessions in video format, but these videos are often kept in individual repositories, and a review of the access logs has shown that employees rarely use them in their day-to-day activities.

To increase efficiency, reduce the load, and gain better insights, this solution looks at how to use generative AI to analyze recorded videos and provide employees with valuable insights relating to their calls. It also supports audio files so you have flexibility around the type of call recordings you use. Generated call transcripts and insights include conversation summary, sentiment, a list of logged actions, and a set of suggested next best actions. These insights are stored in a central repository, unlocking the ability for analytics teams to have a single view of interactions and use the data to formulate better sales and support strategies.

Organizations typically can’t predict their call patterns, so the solution relies on AWS serverless services to scale during busy times. This enables you to keep up with peak demands, but also scale down to reduce costs during times such as seasonal holidays when the sales, engineering, and support teams are away.

This post provides guidance on how you can create a video insights and summarization engine using AWS AI/ML services. We walk through the key components and services needed to build the end-to-end architecture, offering example code snippets and explanations for each critical element that help achieve the core functionality. This approach should enable you to understand the underlying architectural concepts and provides flexibility for you to either integrate these into existing workloads or use them as a foundation to build a new workload.

Solution overview

The following diagram illustrates the pipeline for the video insights and summarization engine.

To enable the video insights solution, the architecture uses a combination of AWS services, including the following:

• Amazon API Gateway is a fully managed service that makes it straightforward for developers to create, publish, maintain, monitor, and secure APIs at scale.
• Amazon Bedrock is 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.
• Amazon DynamoDB is a fully managed NoSQL database service that provides fast and predictable performance with seamless scalability.
• AWS Lambda is an event-driven compute service that lets you run code for virtually any type of application or backend service without provisioning or managing servers. You can invoke Lambda functions from over 200 AWS services and software-as-a-service (SaaS) applications.
• Amazon Simple Storage Service (Amazon S3) is an object storage service offering industry-leading scalability, data availability, security, and performance. You can use Amazon S3 to securely store objects and also serve static websites.
• Amazon Transcribe is an automatic speech recognition (ASR) service that makes it straightforward for developers to add speech-to-text capability to their applications.

For integration between services, we use API Gateway as an event trigger for our Lambda function, and DynamoDB as a highly scalable database to store our customer details. Finally, video or audio files uploaded are stored securely in an S3 bucket.

The end-to-end solution for the video insights and summarization engine starts with the UI. We build a simple static web application hosted in Amazon S3 and deploy an Amazon CloudFront distribution to serve the static website for low latency and high transfer speeds. We use CloudFront origin access control (OAC) to secure Amazon S3 origins and permit access to the designated CloudFront distributions only. With Amazon Cognito, we are able to protect the web application from unauthenticated users.

We use API Gateway as the entry point for real-time communications between the frontend and backend of the video insights and summarization engine, while controlling access using Amazon Cognito as the authorizer. With Lambda integration, we can create a web API with an endpoint to the Lambda function.

To start the workflow, upload a raw video file directly into an S3 bucket with the pre-signed URL given through API Gateway and a Lambda function. The updated video is fed into Amazon Transcribe, which converts the speech of the video into a video transcript in text format. Finally, we use large language models (LLMs) available through Amazon Bedrock to summarize the video transcript and extract insights from the video content.

The solution stores uploaded videos and video transcripts in Amazon S3, which offers durable, highly available, and scalable data storage at a low cost. We also store the video summaries, sentiments, insights, and other workflow metadata in DynamoDB, a NoSQL database service that allows you to quickly keep track of the workflow status and retrieve relevant information from the original video.

We also use Amazon CloudWatch and Amazon EventBridge to monitor every component of the workflow in real time and respond as necessary.

AI/ML workflow

In this post, we focus on the workflow using AWS AI/ML services to generate the summarized content and extract insights from the video transcript.

Starting with the Amazon Transcribe StartTranscriptionJob API, we transcribe the original video stored in Amazon S3 into a JSON file. The following code shows an example of this using Python:

job_args = {
    'TranscriptionJobName': jobId,
    'Media': {'MediaFileUri': media_uri},
    'MediaFormat': media_format,
    'LanguageCode': language_code,
    'Subtitles': {'Formats': ['srt']},
    'OutputBucketName': output_bucket_name,
    'OutputKey': jobId + ".json"
}
if vocabulary_name is not None:
    job_args['Settings'] = {'VocabularyName': vocabulary_name}
response = transcribe_client.start_transcription_job(**job_args)

The following is an example of our workload’s Amazon Transcribe output in JSON format:

{
    "jobName": "a37f0f27-0908-45eb-8d98-8efc3a9d4590-1698392975",
    "accountId": "8469761*****",
    "results": {
        "transcripts": [{
                "transcript": "Thank you for calling, my name is Ivy. Can I have your name?..."}],
        "items": [{
                "start_time": "7.809","end_time": "8.21",
                "alternatives": [{
                        "confidence": "0.998","content": "Thank"}],
                "type": "pronunciation"
            },
            ...
        ]
    },
    "status": "COMPLETED"
}

As the output from Amazon Transcribe is created and stored in Amazon S3, we use Amazon S3 Event Notifications to invoke an event to a Lambda function when the transcription job is finished and a video transcript file object has been created.

In the next step of the workflow, we use LLMs available through Amazon Bedrock. LLMs are neural network-based language models containing hundreds of millions to over a trillion parameters. The ability to generate content has resulted in LLMs being widely utilized for use cases such as text generation, summarization, translation, sentiment analysis, conversational chatbots, and more. For this solution, we use Anthropic’s Claude 3 on Amazon Bedrock to summarize the original text, get the sentiment of the conversation, extract logged actions, and suggest further actions for the sales team. In Amazon Bedrock, you can also use other LLMs for text summarization such as Amazon Titan, Meta Llama 3, and others, which can be invoked using the Amazon Bedrock API.

As shown in the following Python code to summarize the video transcript, you can call the InvokeEndpoint API to invoke the specified Amazon Bedrock model to run inference using the input provided in the request body:

modelId = 'anthropic.claude-3-sonnet-20240229-v1:0'
accept = 'application/json'
contentType = 'application/json'
    
prompt_template = """
The following is the transcript from one of our sales representatives and our customer.
The AI is a tool that the sales representative uses to obtain a brief summary of what the conversation was about. The AI based this summary on the contents of the conversation and does not make up events that did not happen.
     The transcript is:
     <text>
       {}
     </text>
What is the 2 paragraphs summary of the conversation?
"""
    
PROMPT = prompt_template.format(raw_text)
   	
body = json.dumps(
     {
     	"messages": [
            {
              "role": "user",
              "content": [
                 {"type": "text", "text": PROMPT}
              ],
             }
            ],
           "anthropic_version": "bedrock-2023-05-31",
           "max_tokens": 512,
           "temperature": 0.1,
           "top_p": 0.9
        }
    )
response = bedrock.invoke_model(body=body, modelId=modelId, accept=accept, contentType=contentType)
response_body = json.loads(response["body"].read())
summary = response_body["content"][0]["text"]

You can invoke the endpoint with different parameters defined in the payload to impact the text summarization:

• temperature – temperature is used in text generation to control the level of randomness of the output. A lower temperature value results in a more conservative and deterministic output; a higher temperature value encourages more diverse and creative outputs.
• top_p – top_p, also known as nucleus sampling, is another parameter to control the diversity of the summaries text. It indicates the cumulative probability threshold to select the next token during the text generation process. Lower values of top_p result in a narrower selection of tokens with high probabilities, leading to more deterministic outputs. Conversely, higher values of top_p introduce more randomness and diversity into the generated summaries.

Although there’s no universal optimal combination of top_p and temperature for all scenarios, in the preceding code, we demonstrate sample values with high top_p and low temperature in order to generate summaries focused on key information, maintaining fidelity to the original video transcript while still introducing some degree of wording variation.

The following is another example of using the Anthropic’s Claude 3 model through the Amazon Bedrock API to provide suggested actions to sales representatives based on the video transcript:

prompt_template = """
The following is the transcript from one of our sales representatives and our customer.
The AI is a tool that the sales representative uses to look into what additional actions they can use to increase sales after the session. The AI bases the suggested actions on the contents of the conversation and what it thinks might help increase the customers satisfaction and loyalty.

The transcript is:
     <text>
      {}
     </text>

     Using the transcript above, provide a bullet point format for suggested actions the sales representative could do to increase follow on sales.
    """


PROMPT = prompt_template.format(raw_text)
    
body = json.dumps(
   	{
     	"messages": [
         	  {
              "role": "user",
              "content": [
                 {"type": "text", "text": PROMPT}
               ],
             }
            ],
            "anthropic_version": "bedrock-2023-05-31",
            "max_tokens": 1024,
            "temperature": 0.1,
            "top_p": 0.9
        }
    )

response = bedrock.invoke_model(body=body, modelId=modelId, accept=accept, contentType=contentType)
response_body = json.loads(response["body"].read())
suggested_actions = response_body["content"][0]["text"]

After we successfully generate video summaries, sentiments, logged actions, and suggested actions from the original video transcript, we store these insights in a DynamoDB table, which is then updated in the UI through API Gateway.

The following screenshot shows a simple UI for the video insights and summarization engine. The frontend is built on Cloudscape, an open source design system for the cloud. On average, it takes less than 5 minutes and costs no more than $2 to process 1 hour of video, assuming the video’s transcript contains approximately 8,000 words.

Future improvements

The solution in this post shows how you can use AWS services with Amazon Bedrock to build a cost-effective and powerful generative AI application that allows you to analyze video content and extract insights to help teams become more efficient. This solution is just the beginning of the value you can unlock with AWS generative AI and broader ML services.

One example of how this solution could be taken further is to expand the scope to help tackle some of the logged actions from calls. The addition of services such as Amazon Bedrock Agents could help automate some of the responses, such as forwarding relevant documentation like product specifications, price lists, or even a simple recap email. All of these could save effort and time, enabling you to focus more on value-added activities.

Similarly, the centralization of all this data could allow you to create an analytics layer on top of a centralized database to help formulate more effective sales and support strategies. This data is usually lost or misplaced within organizations because people prefer different methods for note collection. The proposed solution gives you the freedom to centralize data but also augment organization data with the voice of the customer. For example, the analytics team could analyze what employees did well in calls that have a positive sentiment and offer training or guidance to help everyone achieve more positive customer interactions.

Conclusion

In this post, we described how to create a solution that ingests video and audio files to create powerful, actionable, and accurate insights that an organization can use through the power of Amazon Bedrock generative AI capabilities on AWS. The insights provided can help reduce the undifferentiated heavy lifting that customer-facing teams encounter, and also provide a centralized dataset of customer conversations that an organization can use to further improve performance.

For further information on how you can use Amazon Bedrock for your workloads, see Amazon Bedrock.

About the Authors

Simone Zucchet is a Solutions Architect Manager at AWS. With over 6 years of experience as a Cloud Architect, Simone enjoys working on innovative projects that help transform the way organizations approach business problems. He helps support large enterprise customers at AWS and is part of the Machine Learning TFC. Outside of his professional life, he enjoys working on cars and photography.

Vu San Ha Huynh is a Solutions Architect at AWS. He has a PhD in computer science and enjoys working on different innovative projects to help support large enterprise customers.

Adam Raffe is a Principal Solutions Architect at AWS. With over 8 years of experience in cloud architecture, Adam helps large enterprise customers solve their business problems using AWS.

Ahmed Raafat is a Principal Solutions Architect at AWS, with 20 years of field experience and a dedicated focus of 6 years within the AWS ecosystem. He specializes in AI/ML solutions. His extensive experience spans various industry verticals, making him a trusted advisor for numerous enterprise customers, helping them seamlessly navigate and accelerate their cloud journey.

Read More