Academy / Generative AI Engineering

Path 3 · Course 1 · For developers and technical teams

Building LLM Applications with RAG

Connect large language models to your own documents: embeddings and vector search, chunking, hybrid retrieval and reranking, grounded answers with citations, and evaluating what you build — with working Python code.

📘 6 modules⏱ About 4 hours✅ Knowledge check per module🎓 Final assessment & certificate🎯 Basic Python

0 of 6 modules complete

Start Module 1 Final assessment

Modules

  1. 1

    Why LLMs need your data

    What LLMs don't know, prompt-stuffing vs RAG vs fine-tuning, the two halves of the RAG pipeline, and where RAG fails.

    25 min
  2. 2

    Embeddings and vector search

    Embeddings, cosine similarity, semantic search in code, vector databases and ANN indexes, and choosing an embedding model.

    35 min
  3. 3

    Preparing documents: chunking and metadata

    Loading and cleaning documents, why and how to chunk, chunking strategies and code, metadata, and keeping the index fresh.

    30 min
  4. 4

    Retrieval that works

    Lexical vs semantic search, hybrid search and reciprocal rank fusion, reranking, contextual retrieval, query rewriting, filters and permissions.

    35 min
  5. 5

    Generating grounded answers

    A grounded prompt template, design choices, an end-to-end mini RAG system in Python, and showing sources to users.

    35 min
  6. 6

    Evaluating and improving RAG

    Retrieval and generation metrics, building a test set, recall@k and MRR in code, LLM-as-judge, the improvement loop, advanced RAG and a production checklist.

    35 min