Machine Learning on AWS (Part 7): Generative AI and Foundation Models

Part 7 of my Machine Learning on AWS series, prepared while studying for the AWS Certified Machine Learning Engineer – Associate (MLA-C01) exam. The first six parts covered models you train yourself. This one covers the other half of the field: foundation models that someone else trained, which you adapt rather than build. In this video: What makes a model a foundation model, and why the workflow inverts Amazon Bedrock, SageMaker JumpStart, and when each one fits Prompt engineering that actually changes the output Retrieval-augmented generation (RAG) explained properly The customisation ladder: prompting, RAG, fine-tuning, continued pre-training Guardrails, evaluation, and why these models fail fluently What drives the bill when you pay per token AWS services covered: Amazon Bedrock, Amazon SageMaker JumpStart, SageMaker endpoints, Bedrock Guardrails