Retrieval-augmented generation, from chunk to cited answer
Retrieval is the LLM system companies ship most, so interviews probe it from every side. This lesson covers when to use it at all, how documents become searchable chunks, how vector indexes and hybrid search work, how permissions and freshness stay correct, and how to tell a retrieval failure from a generation failure.
What you'll be able to do
- Choose between retrieval, long context and fine-tuning for a requirement and defend the choice
- Design the ingest and query paths, including chunking, embeddings, hybrid search, fusion and reranking
- Explain HNSW, IVF and product quantisation and the recall, latency and memory trade each makes
- Keep a retrieval system correct under permissions, document changes and embedding-model migrations
- Evaluate retrieval and generation separately and trace a wrong answer to the stage that caused it
Before this: prompts-and-context
Log in to read the rest of this lesson
Retrieval is the LLM system companies ship most, so interviews probe it from every side. This lesson covers when to use it at all, how documents become searchable chunks, how vector indexes and hybrid search work, how permissions and freshness stay correct, and how to tell a retrieval failure from a generation failure. This walkthrough runs about 18 minutes, with runnable code you can edit and re-run as you read. A new account gets 7 days of Pro free, no card required.
Log in to continueThe first module of every track is free to read on the web: see what's open in AI interviews.