Course map · Season 2 of 15
Technical season, now running
S2 Retrieval and RAG
Grounding models in your own data with embeddings, search, chunking, reranking, and retrieval-augmented generation (RAG).
- 3 lessons live
- 14 on the way
- 7 core ideas
Lessons, in order
-
Season 2: Retrieval and RAG
Season introduction: build an assistant that answers from your own documents, then measure every stage until you know why it is right and why it is wrong.
Rule Build the simplest complete system. Measure it. Change one stage at a time, and keep only what clears a bar set in advance.
-
Why retrieval exists
Retrieval hands the model the right passage at the moment of the question, which fixes stale and private facts but moves the failure to search.
Rule Use retrieval when the answer lives in your own data. Then log what it retrieved, and measure search and answer separately.
-
Your first RAG system
A complete RAG system with the simplest choice at every stage, run on all 520 test questions, shows most wrong answers never reached the model.
Rule Build the simplest complete pipeline and measure it before tuning any stage.
-
Measuring it
-
Chunking
-
Search: keyword, vector, hybrid
-
Rewriting the question
-
Reranking
-
Filters, permissions and fresh documents
-
Building the prompt
-
Grounding and citations
-
Field note: does a citation rule change the answer?
-
Field note: swapping the embedding model
-
Field note: long context or retrieval?
-
Tracing a wrong answer
-
When RAG is the wrong answer
-
Season 2 recap