RAG & Agentic AI Systems

A coach that teaches by asking — not by lecturing

The Lighthouse Coach is an AI Socratic coach built on Cognosco Learning's RAG & Agentic AI Systems course, in the AI at Work category. It does not play videos at you. It holds a conversation: it asks you a question, listens to your answer, names the misconception it hears, affirms what you got right, and asks the next question — the way a great coach works in the room.

A retrieval system makes a checkable promise. An ordinary chatbot answers from what it absorbed when it was built. A retrieval-augmented system answers from these documents, the ones you chose, and shows where each answer came from — and every stage of its pipeline can fail in a way that looks like the same confident wrong answer: the corpus lacked the document, the chunk cut the answer in half, the search ranked a look-alike first, or the model blended the passages it was handed.

Module 1 builds retrieval that finds the answer. State the contract first — documents, scope, refusal, stakes, and the judgment clause; audit the corpus on authority, coverage, duplication and contradiction, and provenance; cut documents to fit their content; understand embeddings and the index as mechanism; measure recall and precision; rerank the shortlist; run exact words and meaning together; and transform the question toward the documents' vocabulary.

Module 2 makes answers verifiable and systems runnable. Ground every claim inside the evidence and cite it at claim level; evaluate retrieval and answers as separate grades that are never merged; put the agent on retrieval duty only, routing and iterating under bounds; keep the corpus fresh and permissioned; and run the whole system inside a token and latency budget computed from its own parts.

It works in two movements. First, Module 1 — Building Retrieval That Finds the Answer — work through Concepts 1–9, finishing with The Retrieval Foundation; then, Module 2 — Answers You Can Verify, Systems You Can Run — work through Concepts 10–18, finishing with The System in Production.

Both capstones produce real artifacts — The Retrieval Foundation, a foundation design for Fennick Parts Co.'s retrieval system that a build team could work from; and The System in Production, an operating plan for running that system responsibly — each built on the running case and again on your own live retrieval project.

One workplace, worked honestly, the whole way through. Fennick Parts Co., a fictional distributor of industrial components, is building a retrieval system over its own documentation — product manuals, warranty terms, and safety bulletins — and makes, in these pages, most of the mistakes real teams make, in an order that lets you learn from each one cheaply. Every worked number recomputes exactly from the figures given on its own page.

Scope and honesty are stated up front. This book owns retrieval systems: general agent construction belongs to a companion course, and agents appear here only as controllers of retrieval. Product facts are labeled as product facts — wherever the course touches one, it says so in place: as of this writing, check the current documentation.

The course is self-paced and always on — roughly 30 minutes per concept, about 9 hours in full, taken in as many sittings as you like. You can type your answers, or talk out loud: the Voice Coach speaks its questions and listens to your spoken replies. Learners who already know the material can test out concept by concept.

This is the two hundred and third course in The Lighthouse Signature Series — the online counterpart to the one-day, in-person learning Cognosco Learning has delivered since 2002.