Optimizing Foundation Models
AWS Skill Builder
Last updated February 23, 2026
Optimizing Foundation Models is a free, one-hour course on two of the main ways to get better results from a foundation model: retrieval-augmented generation (RAG) and fine-tuning. On the RAG side, it looks at the AWS services that store the embeddings a model retrieves against — the vector-database piece — and at how agents handle multi-step tasks. On the fine-tuning side, it covers the methods available and, just as important, how to prepare the data a fine-tuning job depends on. Throughout, it frames each technique against a business case and how to judge whether the result actually meets the objective. The material is conceptual, delivered through interactive elements, text, and illustrative graphics.
What you'll learn
- What RAG is, and the role of vector databases
- The AWS services used to store embeddings
- The role of agents in multi-step tasks
- Methods for fine-tuning a foundation model
- How to prepare data for a fine-tuning job
- Evaluating whether an FM meets its business objective
Frequently asked questions about Optimizing Foundation Models
Who is Optimizing Foundation Models for?
Anyone interested in AI and machine learning, independent of job role, who wants to understand how RAG and fine-tuning are used to improve a foundation model's performance.
Is Optimizing Foundation Models free?
Yes — Optimizing Foundation Models is completely free to take.
What are the prerequisites for Optimizing Foundation Models?
Recommends completing a foundational ML and AI course and an AI use-cases course first.
Why we suggest this course
A focused look at the two workhorse techniques for improving a foundation model — RAG and fine-tuning — including the often-skipped question of how to prepare data for fine-tuning. One thing to know: it is built by AWS and names specific AWS services for the storage and retrieval pieces (OpenSearch, DynamoDB, RDS, S3, Kinesis), so expect the platform framing — the RAG and fine-tuning concepts themselves are general.