Chunking, embeddings, vector search, and fixing hallucinations in production.
Design retrieval systems that actually reduce hallucination — chunking, embeddings, hybrid search, re-ranking, and the production concerns most tutorials skip.
Engineers building any product where the model has to answer from your data — not just its training data.
The guides are numbered — read in order for the curriculum path, or jump straight to the one you need. Each card is self-contained.
Grounding LLMs in your own data; the open-book analogy.
How meaning-based retrieval beats keyword search.
Storing/searching vectors; comparing the main options.
Chunk size, overlap, structure-aware splitting.
The indexing + retrieval-generation flow end to end.
Tuning what gets fetched and how much.
Diagnosing retrieval vs generation failures.
Retrieve-broad-then-rerank for precision.
Catching exact terms, codes and names embeddings miss.
Reformulation and multi-hop questions.
Letting the model decide whether/what/how to retrieve.
Measuring retrieval and answer quality objectively.
Operating a RAG system reliably at scale.