Build a Production RAG from Scratch
Go from raw documents to a deployed retrieval-augmented generation system — chunking, embeddings, hybrid search, and evaluation, built the way production teams actually do it.
Curriculum
Build the ingestion pipeline that turns raw documents into clean, retrievable chunks.
Generate embeddings and store them in a vector database ready for retrieval.
Combine keyword and vector retrieval for higher-accuracy search results.
Wire retrieval into generation and build an evaluation harness to measure quality.
Package the full pipeline as a deployed API you can query end to end.
What you'll build
This course follows one complete project from start to finish — taking raw documents through chunking, embedding, hybrid retrieval, generation, and evaluation, ending with a deployed API you can query end to end.
Every project inside this course is a checkpoint toward that final system — you can inspect, run, and modify each one independently.
One subscription. Every course. Real projects.
Included with an Analog Data subscription. Includes the course, linked projects, source code, and future updates.
FAQ
The course shows you how to use hosted models (which need an API key) and open-weight models you can run locally. You can complete the projects either way.
Your Analog Data subscription unlocks this course, every other course, and the linked projects as they are published. It is the only way to access the course library.
You need basic Python. The course introduces embeddings, retrieval, and evaluation from first principles — no prior ML background required.
Stop reading. Start retrieving.
Real projects. A production RAG system you can explain, deploy, and defend.