What you’ll learn
- Prepare a permitted document collection and searchable chunks.
- Connect retrieval to answers with traceable source references.
- Evaluate retrieval misses, unsupported answers and access boundaries.
Proposed course content
6 sections · 13 illustrative lesson records
Documents and permissions3 lessons
- Introduction to RAG & Knowledge-Based Assistants
Build retrieval-augmented assistants with permitted documents, transparent citations, uncertainty handling and reproducible retrieval evaluation.
Public introduction · Video not supplied - Record source provenance and access rules
Record source provenance and access rules. Module: Documents and permissions.
Enroll to access · Video not supplied - Normalize content and handle unsupported formats
Normalize content and handle unsupported formats. Module: Documents and permissions.
Enroll to access · Video not supplied
Chunking and indexing2 lessons
- Compare chunk boundaries and metadata choices
Compare chunk boundaries and metadata choices. Module: Chunking and indexing.
Enroll to access · Video not supplied - Build a small reproducible search index
Build a small reproducible search index. Module: Chunking and indexing.
Enroll to access · Video not supplied
Retrieval methods2 lessons
- Compare keyword and embedding retrieval
Compare keyword and embedding retrieval. Module: Retrieval methods.
Enroll to access · Video not supplied - Inspect ranking and query failures
Inspect ranking and query failures. Module: Retrieval methods.
Enroll to access · Video not supplied
Grounded generation2 lessons
- Supply source context and traceable citations
Supply source context and traceable citations. Module: Grounded generation.
Enroll to access · Video not supplied - Abstain when evidence is missing or conflicting
Abstain when evidence is missing or conflicting. Module: Grounded generation.
Enroll to access · Video not supplied
Evaluation2 lessons
- Create synthetic questions with expected evidence
Create synthetic questions with expected evidence. Module: Evaluation.
Enroll to access · Video not supplied - Measure retrieval coverage and unsupported answer rates
Measure retrieval coverage and unsupported answer rates. Module: Evaluation.
Enroll to access · Video not supplied
Application safeguards2 lessons
- Enforce document access boundaries and data retention choices
Enforce document access boundaries and data retention choices. Module: Application safeguards.
Enroll to access · Video not supplied - Record safe diagnostics and refresh the index consistently
Record safe diagnostics and refresh the index consistently. Module: Application safeguards.
Enroll to access · Video not supplied
Requirements
- Basic Python programming and model/API familiarity.
- Licensed, owned or synthetic documents appropriate for the chosen model environment.
Tools and resources
- A development computer with resources appropriate to the proposed local exercises.
- Python, text-processing tools and a local search or vector index.
- A permitted model interface, synthetic documents and evaluation scripts.
Description
Build retrieval-augmented assistants with permitted documents, transparent citations, uncertainty handling and reproducible retrieval evaluation.
Who this course is for
Developers building evidence-grounded document assistants.
Example practical project
Proposed capstone: an assistant over synthetic technical notes with source-linked answers, abstention cases and retrieval evaluation.