Applications of Foundation Models
The largest exam domain. Choosing a foundation model, RAG and vector databases, prompt engineering, fine-tuning, and how to evaluate what you built.
Knowing what a foundation model is does not tell you how to build something useful with one. This is where the course turns from concepts to construction: you have a model, a business problem, and a set of choices about how to connect the two. Get those choices wrong and you overpay, leak private data into a prompt, or ship a system nobody can trust.
This is the heaviest domain on the exam, and it rewards judgment over memorization. It walks the full path of an application: pick the right model, ground it in your own data with RAG, shape its behavior with prompts, adapt it through fine-tuning when prompts are not enough, and then measure whether the result actually works.
What This Domain Covers
- choosing a foundation model and tuning inference parameters like temperature and top-p
- Retrieval Augmented Generation (RAG) and the vector databases on AWS that make it work
- the customization spectrum from prompting to fine-tuning, and the tradeoffs at each step
- AI agents inside applications, and where they fit alongside RAG
- prompt engineering: anatomy, techniques, best practices, and prompt attacks like injection
- how foundation models are trained, the main fine-tuning methods, and preparing data for them
- evaluating models with the right metrics, plus how to assess RAG and agent applications against business goals
Why It Matters
At 28 percent of the exam, this domain carries more weight than any other, and its questions are scenario-based rather than definitional. You will be handed a situation, a constraint, and four plausible options, then asked which approach fits. That is exactly the skill this domain builds: when to reach for RAG instead of fine-tuning, when a better prompt beats a bigger model, and how to tell whether what you shipped is good enough.
The payoff outlasts the exam. These are the same decisions real teams make when they put a foundation model into production, and getting them right is the difference between a demo and a product people can rely on.