SUSHANT GUNDLA · LEARN← the site
← All courses

A hands-on course · 12 lessons

RAG

Retrieval that actually retrieves.

Your answer can never be better than what you retrieved. Embeddings, vector stores, chunking, hybrid search and reranking — the whole pipeline, and how to tell which part of it is letting you down.

Lessons run about five to eight minutes each and end with a short recall quiz, so you find out what stuck before you move on.

Part 1Embeddings

01

What an embedding really is

Text as a point in space, and an honest list of what that buys you and what it costs you.

~7 min
02

Choosing an embedding model

The five axes that decide it, and why the leaderboard is a filter rather than an answer.

~7 min
03

Chunking

Cutting documents into the units you actually retrieve, and the header line that fixes most bad retrieval.

~8 min

Part 2Storing vectors

04

Vector databases

What one actually is, and the ladder of options you should climb only as far as you need to.

~6 min
05

Indexes and the trade you are making

Approximate search, the knobs that control it, and how to prove you tuned it well enough.

~8 min
06

Metadata and filtering

The half of retrieval that is not semantic, including the half that is a security boundary.

~7 min

Part 3Retrieval

07

Hybrid search

Run keyword search next to vector search and fuse the two ranked lists, so exact strings stop falling through the cracks.

~7 min
08

Reranking

Retrieve fifty candidates cheaply, then have a slower model read query and passage together and pick the best five.

~7 min
09

Query rewriting

Fix the query before you search - expand it, split it, turn it into a fake answer, or make a follow-up stand on its own.

~8 min

Part 4Proving it works

10

Measuring retrieval

Build a fifty-query eval set by hand, score recall@k, MRR and nDCG, and always test retrieval separately from generation.

~8 min
11

Citations and grounding

Give every chunk an id, require the answer to cite it, then check the quoted span really appears where the model said it did.

~7 min
12

When RAG is the wrong tool

Recognise the questions retrieval structurally cannot answer, and pick the tool that can.

~7 min

Work through it in order the first time — each lesson leans on the one before it. After that, come back to any of them on their own.