Advanced · module A9
Retrieval-Augmented Generation and Knowledge Systems
By the end: Build, evaluate and secure a production-grade retrieval-augmented system
- Covers
- Grounding LLMs in your own data, from vector search to production RAG
- Contents
- 2 chapters, 11 topics
- Time
- About 3 hours
- Level
- Advanced
- Your progress
- 0 / 11
This module builds on Generative AI and Large Language Models in Practice. We suggest looking at it first, but the choice is yours.
Chapter 1 Retrieval
- 1.110 minBeing written
RAG Architecture
Ingestion, indexing, retrieval and generation
Interactive: trace a query through RAG
- 1.212 minBeing written
Document Ingestion and Chunking
Parsing files, chunk sizes, overlap and metadata
Code: chunk a document set
- 1.315 minBeing written
Embedding Models and Vector Databases
Choosing embeddings, and HNSW and IVF indexes
Code: a vector index with FAISS
- 1.415 minBeing written
Hybrid Search and Reranking
BM25 combined with vectors, and cross-encoder reranking
Code: add reranking
- 1.510 minBeing written
Query Rewriting and Routing
Multi-query, HyDE and routing to sources
Code: query rewriting
Chapter 2 Generation and Quality
- 2.112 minBeing written
Grounded Generation and Citations
Answering only from sources, with citations
Code: answers with citations
- 2.215 minBeing written
Knowledge Graphs and GraphRAG
Entities, relations and graph-based retrieval
Code: a small knowledge graph
- 2.315 minBeing written
Evaluating RAG Systems
Retrieval recall, faithfulness and answer relevance
Code: an evaluation set
- 2.412 minBeing written
Securing RAG
Access control, prompt injection through documents and data leakage
Lab: attack and defend
- 2.545 minBeing written
Project: A Production RAG Application
Ingestion, search, generation, evaluation and a user interface
Project
- 2.620 minBeing written
Module Review and Assessment
Coding tasks and a scored quiz
Graded project