1

A Gentle Introduction to Generative AI and AI Agents on AWS

Imagine that you are playing a video game where you can create and customize your own set of characters. You can think of generative AI as a creative assistant that automatically suggests unique abilities, costumes, and names for your custom characters. Given that generative AI is a subset of machine learning, we can then think of machine learning (ML) as the game engine that powers not only the character customization feature but also every aspect of the game's mechanics, learning, and evolution based on the player interaction data. Continuing this analogy, machine learning engineering (ML engineering) is like designing, building, and maintaining the game engine to ensure it continuously learns from player data to improve the gaming experience. AI agents takes this a step further by acting like autonomous players in the game who can make decisions, perform tasks, and interact with the environment based on what they have learned.

Before we dive deeper into machine learning engineering, it is essential that we have a good grasp of what generative AI is, its use cases, and the concepts relevant to AI agents. This introductory chapter will help us learn several concepts that will be important for the succeeding chapters of this book. One of the fastest ways to learn new concepts is to learn by doing. In this chapter, we will explore various examples using a managed service called Amazon Bedrock that will help us understand generative AI and the principles behind AI agents. We will also build an AI agent using Strands Agents to help you better understand how these concepts come together in practice, demonstrating how agents can reason, interact with models, and perform tasks autonomously.

We will cover the following topics in this chapter:

• Generative AI for the modern machine learning engineer
• Exploring foundation models in Amazon Bedrock
• Setting up and configuring your SageMaker Studio environment
• Configuring IAM permissions for your SageMaker Studio Space
• Introduction to AI agents with Amazon Bedrock and Strands Agents

Do not worry if this is your first time encountering some of these terms, concepts, and solutions, as we will go through each one in more detail later in this chapter. With this in mind, let's get started!