104Course · Foundation
Retrieval-Augmented Generation
The pattern that made LLMs useful.
The pattern that made LLMs useful in production. Embeddings, vector stores, chunking, hybrid search, reranking — all taught with the honest math of where each helps.
The Scholar builds a real RAG system over a corpus they know, and measures it.
- 01
Explain RAG at a working level.
- 02
Design a retrieval pipeline.
- 03
Evaluate a RAG system's answers.
- 04
Choose between fine-tuning and retrieval defensibly.
Prerequisites
- — Prompt Engineering.
Pacing
~5 hours including the Lab.
- IModule
The pattern
Retrieve, then generate.
- IIModule
Retrieval infrastructure
The catalogue.
- IIIModule
Retrieval pipelines
Beyond the naive lookup.
A Note from the Faculty
Every serious AI product ships either RAG or a good reason not to.