RAG-301

Production RAG Systems

RAG is not a tutorial-grade architecture; it is a real distributed system with real failure modes. This is the course for treating it like one.

ABOUT THIS COURSE

What you will learn.

Most production RAG systems are bad in the same five ways. They retrieve too much, they retrieve the wrong things, they don’t notice when retrieval failed, they bill the user too much, and they hallucinate confidently when the right answer wasn’t in the index. None of these failures is a model problem. All of them are engineering problems.

This six-week course is the rigorous treatment. We cover ingestion and chunking as the load-bearing decisions they are, embedding choice and evaluation, hybrid and multi-stage retrieval, reranking, query rewriting, citation and grounding strategies, the evaluation discipline that catches RAG regressions, and the operational practices that keep an index honest as documents change. The capstone is a working RAG service with a real eval suite, deployed against a corpus of your choice.

This is the most-requested course in the Solutech catalog. We run it monthly, and it routinely sells out two weeks ahead.

WHAT YOU’LL BUILD

Four substantial projects.

Project 01

An ingestion pipeline you trust

Build a pipeline that handles real-world documents — messy ones — without losing structure.

Project 02

A hybrid retrieval system with reranking

Combine lexical and semantic retrieval with a reranker that earns its inference cost.

Project 03

A grounded answer pipeline with citations

Build an answer pipeline that cites correctly and admits when it does not know.

Project 04

A RAG eval suite

Construct retrieval-level, generation-level, and end-to-end evaluations that disagree usefully.

CURRICULUM

Week by week.

FIT

Who this is for — and who it is not.

For you if

Engineers shipping or about to ship a RAG-backed product.
Teams whose first RAG pilot worked in demo and is now disappointing real users.
Senior engineers being asked to “add search to our LLM thing.”

Probably not for you if

Engineers without LLM experience — start with AI-101.
Researchers focused on retrieval model training — this is engineering.
People looking for a one-evening tutorial — RAG is a six-week subject.

YOUR INSTRUCTOR

Taught by an operator.

Research Lead

Yuki Tanaka

Yuki led the retrieval-quality group at DeepMind before joining Solutech full time. She co-authored four widely-cited papers on dense passage retrieval and hybrid lexical-semantic search, and her course-week on reranking has become required reading inside two well-known AI startups. Her teaching style is direct: she will tell you, on day one, which of your assumptions about RAG are wrong, and she will be right.

FAQ

Questions we’re asked often.

RAG-301 · Next cohort starts soon