A free course · 54 concepts · 12 chapters
RAG that works on real documents.
Learn retrieval augmented generation one concept at a time, from the first principles to the problems you only meet in production. Every idea is tested on the same documents, so you see what each change is worth.
- 1Load your documents
- 2Cut them into chunks
- 3Turn chunks into numbers
- 4Find the right pieces
- 5Build the prompt
- 6Answer with sources
Walk through all six steps, with a live search →
Measured
Numbers, not opinions
Every technique is tested on the same 25 documents and 47 questions. You see what moved and by how much.
Real labs
Retrieval that actually runs
Chunking, keyword search, embeddings and metrics compute in your browser. No sliders over fake scores.
Production
How it breaks, and the fix
Stale answers, split tables, leaked documents, injected instructions. Each page names its failure mode.
Contents
From first principles to production
Chapter 1
What RAG is, and why
4 concepts
Chapter 2
Tokens, cost and your first RAG
2 concepts
Chapter 3
Getting documents in
6 concepts
Chapter 4
Chunking
6 concepts
Chapter 5
Embeddings
4 concepts
Chapter 6
Vector stores and scores
3 concepts
Chapter 7
Finding the right chunks
6 concepts
Chapter 8
When retrieval fails
1 concepts
Chapter 9
Building the answer
6 concepts
Chapter 10
Measuring
6 concepts
Chapter 11
Safety and operations
6 concepts
Chapter 12
Scale and advanced RAG
4 concepts