RAG for Developers: Build, Diagnose, Ship

This course contains the use of artificial intelligence.

Retrieval-augmented generation (RAG) is how you get a large language model to answer from your own documents instead of guessing from what it memorized in training, and building one that actually works in production is a different skill from a weekend demo.

This course takes you from "what is RAG" to a working, evaluated, production-hardened retrieval-augmented generation system. You'll build a full indexing pipeline (chunking, embeddings, vector database selection, and a hands-on vendor comparison across the real tradeoffs between vector stores), implement hybrid search and reranking, apply advanced retrieval patterns (RAG-Fusion, corrective and adaptive RAG, agentic and graph-based retrieval, and text-to-SQL for structured data), and, unlike almost every other RAG course, learn to evaluate and diagnose your system with RAGAS-style metrics, a structured failure catalog, and a continuous evaluation pipeline that catches regressions automatically, rather than treating "it looks right" as good enough.

A hands-on lesson early on builds a complete minimal RAG pipeline line by line, and a two-part capstone at the end of the core course ties everything together on your own document set, including a dedicated lesson on engineering prompts for grounded, citation-backed answers.

A optional module on production hardening (not required to finish the core course, but included) covers five things most courses skip entirely: defending a RAG system against prompt injection and permission leakage so retrieval never leaks content a user shouldn't see, parsing real-world documents (tables, scans, images) instead of assuming clean text, choosing an embedding model and controlling embedding cost at scale through caching, knowing when a long-context model makes retrieval unnecessary versus when it doesn't, and isolating tenant data correctly in a multi-tenant RAG deployment. This module tracks a fast-moving part of the ecosystem more closely than the rest of the course, so it's the one most likely to see a refresh pass between rounds of edits.

This course is framework-agnostic: concepts are taught with code on screen, not tied to one version of a specific library, so what you learn keeps working after the next API changes.

What's not in this course: fine-tuning language models, building your own embedding model from scratch, and UI/frontend chat interfaces: those are separate skills for a separate course.