Building an 8-way AWS Architect Agent App with Amazon Bedrock

Introduction

• The speakers are Banjo Biami, a Senior Developer Advocate at AWS, and Shar, a software engineer and YouTuber.
• The session focuses on building an agentic workflow using Amazon Bedrock, a fully managed serverless service for building and scaling generative AI applications.

AWS Generative AI Stack

• The AWS generative AI stack consists of various services and tools for building, training, and deploying AI models.
• At the bottom, there are services like Amazon SageMaker, AWS Trinium, and AWS Inferentia for infrastructure and model development.
• Amazon Bedrock is a key service that allows building and scaling generative AI applications using foundation models.
• Other tools in the stack include Amazon Lex (business-focused) and Amazon Q (developer-focused).

Agentic Workflows

• Agentic workflows are about building AI agents that can perceive, decide, and act autonomously to achieve goals, adapting their behavior based on experiences and interactions.
• Key design patterns for agentic workflows include reflection, tool use, planning, and multi-agent collaboration.
• Building agentic systems can help with breaking down complex tasks, increasing autonomy, improving efficiencies, and achieving better scalability.

Building the AWS Architect Agent App

• The speakers demonstrate building an AWS Architect Agent app using Amazon Bedrock.

• The agent is designed to help customers with the following tasks:

  1. Querying the AWS Well-Architected Framework knowledge base
  2. Generating AWS Lambda functions
  3. Describing AWS diagrams
  4. Drawing AWS diagrams
  5. Extracting text from websites

• The agent uses various action groups (tools) to accomplish these tasks, and a knowledge base with the AWS Well-Architected Framework.

• The agent demonstrates its capabilities by:

  • Providing a response to a query about cloud security best practices using the knowledge base
  • Describing an existing AWS diagram for a media streaming solution
  • Creating a new diagram for an e-commerce architecture
  • Generating a AWS CDK script to deploy the e-commerce architecture
  • Creating a Lambda function for sentiment analysis and providing the client code to use it

Key Takeaways

• Agentic workflows empower builders to integrate AI into their personal and professional workflows, leveraging tools like Amazon Bedrock and Amazon Q.
• The best way to utilize these tools is to experiment and find the use cases that work best for your specific needs, as there is no one-size-fits-all solution.
• The workshop "Building Agentic Workflows on AWS" is recommended for a step-by-step guide on creating agentic workflows using Amazon Bedrock and other AWS services.