Building Serverless Apps with AI Services on AWS

Learn how to build a serverless application that can transcribe video, translate subtitles, and generate a title and video description using generative AI.

Introduction

• Serverless apps with AI services on AWS
• What you should know
• What you are going to build in this course

1. Introduction to AI

• Introduction to AI
• Introduction to Amazon AI Services
• Introduction to Gen AI
• Introduction to Amazon Bedrock
• Challenges when building AI and Gen AI applications

2. Getting started with Serverless

• What is serverless?
• Why serverless is great to build AI apps
• Using Amazon Bedrock from serverless services

3. Infrastructure as code

• The important role of infrastructure as code
• Introducing AWS SAM
• Installing and configuring AWS SAM
• Getting started with the project

4. Introduction to Step Functions

• Understanding Step Functions
• Creating your first workflow in the AWS console
• Creating a workflow as IaC
• Introduction to ASL
• Introduction to flow states
• Step Functions integrations
• Modifying the workflow with simple integration
• Challenge: Modify the state machine to add an integration to AI service
• Solution: Modify the state machine to add an integration to AI service

5. Building a Workflow with AI

• Modifying the existing state machine to use Amazon Bedrock
• Challenge: Integrate with Amazon Bedrock
• Solution: Integrate with Amazon Bedrock
• What is event-driven architecture?
• How to trigger a state machine when there is a new S3 file
• How to send an email from the state machine
• Challenge: Integrate to EventBridge event bus to send an event
• Solution: Integrate to EventBridge event bus to send an event

6. Build a Chatbot with GenAI

• What is an Amazon Bedrock agent?
• What you are going to build
• Creating a Lambda function for the agent
• Creating an Open API for the function
• What is an Amazon Bedrock knowledge base?
• Creating a knowledge base
• Creating an Amazon Bedrock agent
• Challenge: Add another function to the agent
• Solution: Add another function to the agent

Conclusion

• What we built
• Cleaning up the environment
• Next steps