Sign in Sign up

Learning path

Retrieval-Augmented Generation (RAG)

Why RAG, chunking, embeddings and vector databases, retrieval, re-ranking, citations and evaluation, with animated films and runnable Python. Part of the Gen AI Developer premium track.

Start with lesson 1
  1. 1

    Why RAG: Answers from Your Own Data

    Retrieval-Augmented Generation explained: why LLMs need your documents, the retrieve-then-generate flow, RAG vs fine-tuning vs long context, and a complete mini RAG system in plain Python.

    4 min read
  2. 2

    Chunking Documents for RAG

    Split PDFs and web pages into chunks that retrieve well: fixed-size with overlap, sentence and heading-aware chunking, metadata for citations, and how chunk size affects answers.

    3 min read
  3. 3

    Embeddings and Vector Databases

    Turn chunks into embeddings with an embedding model, store them in a vector database (Chroma, FAISS, pgvector, Pinecone), and run fast similarity search, with a NumPy vector store…

    3 min read
  4. 4

    Retrieval, Re-ranking and Citations

    Build the answering half of RAG: top-k retrieval, metadata filters, re-ranking, the grounded prompt, citations, and checking that every cited source really supports the answer.

    4 min read
  5. 5

    Retrieval, Re-ranking and Citations

    Build the answering half of RAG: top-k retrieval, metadata filters, re-ranking, the grounded prompt, citations, and checking that every cited source really supports the answer.

    4 min read
  6. 6

    Evaluating a RAG App

    Measure RAG quality properly: a test set of questions, retrieval metrics (hit rate, MRR), answer metrics (faithfulness, correctness), LLM-as-judge, and a regression check before every change.

    4 min read