Generative AI on AWS

Chapter 12. Amazon Bedrock: Managed Service for Generative AI

Throughout the book, you have seen examples of Amazon SageMaker JumpStart for fine-tuning and deploying foundation models using SageMaker infrastructure. Amazon Bedrock, on the other hand, is a managed service that offers a completely serverless experience through a simple API.

In this chapter, you will explore Amazon Bedrock, including how to access the Bedrock API, the available foundation models (FMs), and Bedrock data privacy and network security. You will learn how to use Bedrock to implement retrieval-augmented generation, semantic-search, and agent-based use cases. You will also see how you can privately fine-tune the Bedrock foundation models using your own custom datasets.

First, let’s discuss the available foundation models within Amazon Bedrock—and how to build upon those foundation models.

Bedrock Foundation Models

Amazon Bedrock supports foundation models from Amazon and various third-party companies, including AI21 Labs, Anthropic, Cohere, Meta, Stability AI, and others.

You access these foundation models through the AWS Management Console, AWS CLI, or AWS SDK. The code examples in this chapter will use the AWS SDK for Python called boto3. You can use the Bedrock Python function list_​founda⁠tio⁠nal​_models() to see the most up-to-date list of available models.

Working with Amazon Bedrock is as simple as selecting a foundation model for your use case and then making a few API calls. You can use the Bedrock ...