VFlash: Fast, Accurate Visual Planning via Flash Primitive Decoding
u_Unrolled
D8: RAG & Knowledge Systems
Phase 5: Production ML, RAG, Agents & Evaluation/DOMAIN 8

Domain 8: Retrieval-Augmented Generation & Knowledge Systems

AdvancedUnrolled Pro

Retrieval-augmented generation end to end: BM25 and analyzers, embeddings, document ingestion and chunking, vector databases, hybrid search, reranking and RAG evaluation.

Estimated study time: about 100 hours
Access: Unrolled Pro
Unrolled Pro Domain

Unlock all 8 lessons with Unrolled Pro

Full production code, quantitative capacity math, failure recovery playbooks, and interview coaching.

Get Pro from ₹167/mo (~$2/mo)14-day refund guarantee

Lessons

8 lessons
01

Retrieval I — The Pipeline, the Inverted Index and BM25 from First Principles

Pro 25m

Retrieval-augmented generation (RAG) is the pattern behind almost every assistant that answers from a company's own knowledge: search first, then let the model answer from what was found, citing it.

✓ Runnable code✓ Exercise with solutionStart Lesson
02

Retrieval II — Analyzers, Speed, Sharding, Permissions and Debugging

Pro 35m

Lesson I built retrieval that works on a toy corpus.

✓ Runnable code✓ Exercise with solutionStart Lesson
03

Embeddings — Turning Meaning into Geometry

Pro 20m

An embedding maps text to a vector so that texts with similar meaning land close together. That one idea fixes BM25's biggest weakness, powers semantic search and recommendations, and underlies every vector database.

✓ Runnable code✓ Exercise with solutionStart Lesson
04

Document Ingestion — From Messy Files to Clean, Retrievable Chunks

Pro 20m

"Garbage in, garbage out" is truer for RAG than almost anywhere else. If a navigation menu gets indexed, it matches every query. If a fact is split across two chunks, neither chunk answers the question.

✓ Runnable code✓ Exercise with solutionStart Lesson
05

Vector Databases — Finding Nearest Neighbours Among Millions

Pro 20m

Finding the nearest vectors exactly means comparing the query with every stored vector. That's fine for thousands of vectors and hopeless for hundreds of millions.

✓ Runnable code✓ Exercise with solutionStart Lesson
06

Advanced RAG — Hybrid Search, Reranking, Context and Citations

Pro 20m

"Basic RAG" — embed the question, take the top 5 chunks, generate — is a demo, not a product.

✓ Runnable code✓ Exercise with solutionStart Lesson
07

RAG Evaluation — Measuring Retrieval and Answers Separately

Pro 20m

A wrong RAG answer has two possible causes: retrieval never found the right passage, or the model had it and still got the answer wrong.

✓ Runnable code✓ Exercise with solutionStart Lesson
08

Bridge: From Answering Questions to Taking Actions

Pro 10m

Atlas can now find the right passage and answer from it, with citations and measured quality. But students don't only ask questions.

✓ Runnable codeStart Lesson
Domain 8: RAG & Knowledge Systems · Unrolled