Deploying and Governing AI on AWS

This unit covers selecting appropriate machine learning techniques and AWS services for real-world use cases, designing prompts for LLMs and Bedrock agents, and deploying models for inference with automated MLOps practices. Students will also learn to monitor, evaluate, and apply guardrails (including SageMaker Clarify) to ensure model fairness, safety, and reliability, building on prior foundational concepts and preparing for advanced production workflows in the final unit.

Learning Objectives

• Analyze and justify the selection of ML techniques and AWS ML services for given use cases, considering data types (structured, unstructured, and multi-modal) and model capabilities
• Apply prompt engineering best practices to design and test prompts for Amazon Bedrock agents and LLMs to improve relevance, safety, and task performance
• Demonstrate end-to-end deployment and inference workflows using Amazon SageMaker (including SageMaker Inference) and implement basic MLOps automation for training, deployment, and monitoring
• Evaluate model fairness, bias, and safety using SageMaker Clarify and Bedrock guardrails, and propose remediation or mitigation strategies based on observed metrics