Generative AI with Amazon Bedrock

This unit covers advanced practical skills for using generative and multimodal foundational models via Amazon Bedrock, including image generation with Stable Diffusion, parameter tuning, and building end-to-end text-to-SQL systems. Students will also implement guardrails, maintain chatbot context, and explore model management with SageMaker JumpStart to prepare for deployment and evaluation in the final unit.

Learning Objectives

• Analyze the effects of model parameters and multimodal context strategies by comparing outputs from Stable Diffusion and Bedrock multimodal models on specified prompts and inputs.
• Apply guardrail techniques and Bedrock safety tools to implement content filtering, prompt constraints, and logging within a generative AI pipeline.
• Demonstrate the design and implementation of an end-to-end text-to-SQL system using Amazon Bedrock connected to a sample database, including validation of query accuracy and performance.
• Evaluate model management and deployment options using Amazon SageMaker JumpStart and propose strategies to maintain conversational context and consistent model behavior across sessions.