Why Normal Chat Is Not the Same as Knowledge Retrieval
By the end of this module you can explain why a chat model answers confidently from nothing, and what retrieval changes about that - and does not.
Introduction to RAG and Knowledge Assistants
A knowledge assistant design brief with source rules, metadata, citation policy, missing-answer behavior, permissions, and an evaluation set.
Documents, Chunks, Metadata, and Permissions
By the end of this module you can split documents into chunks that stand alone, attach the metadata that makes an answer citable, and carry permissions at the chunk level.
Embeddings, Indexes, and Retrieval in Plain English
By the end of this module you can explain embeddings, indexes and retrieval in plain English, and say why semantic similarity is poor at exact constraints like a date or a country.
Query Rewriting and Retrieval Quality
By the end of this module you can improve retrieval by fixing the query rather than the model, and measure whether a rewrite actually retrieved better documents.
Grounded Answering with Citations
By the end of this module you can make an assistant answer from retrieved text with citations a reader can follow back to the source record.
Missing Answers and Refusal Behavior
By the end of this module you can design what happens when the answer is not in the documents - a clear refusal beats a fluent guess, and it has to be built deliberately.
Hallucination Risks and Source Conflicts
By the end of this module you can spot hallucination and source conflict in a grounded answer, and decide what the assistant should do when two documents disagree.
Evaluation Sets for RAG
By the end of this module you can build an evaluation set for a RAG system, testing retrieval and answer quality separately so you know which half is failing.
Privacy, Access Control, and Document Lifecycle
By the end of this module you can enforce access control in retrieval rather than in the prompt, and plan for documents that are updated, superseded or deleted.
Designing a Knowledge Assistant Workflow
By the end of this module you can design a knowledge assistant as a whole workflow - ingestion, retrieval, answering, refusal and review - rather than as a single clever prompt.
Capstone: RAG Design Brief and Evaluation Plan
In this capstone you will write a RAG design brief with its chunking and permission model, its refusal behaviour, and an evaluation plan someone else could run against it.
First-release boundary
This is a text-first course shell. Leonardo/image creation, image QC, dependency-heavy runtime QA, external review, and learner pilot evidence remain separate later quality steps.