AI Knowledge Assistants and Retrieval-Augmented Generation

About Course

Design and evaluate a knowledge assistant that retrieves relevant authorised evidence before drafting an answer. Cover source preparation, chunking, lexical and vector search, citations, abstention, access controls, refresh and evaluation. Includes 21 lessons, fictional policy cases, an optional local retrieval lab and a 42-question final quiz.

LO01: Specify a knowledge-assistant scope and an evidence-based test set.
LO02: Create traceable retrieval units that preserve meaning, version and access.
LO03: Explain and test retrieval choices using explicit relevance criteria.
LO04: Retrieve only authorised applicable content and prepare an answer contract.
LO05: Check that answers are supported and do not treat retrieved text as authority.
LO06: Maintain correct meaning and freshness across multi-step knowledge use.
LO07: Evaluate an end-to-end assistant with traceable answers and explicit limitations.

Course Content

Knowledge purpose and evidence
Define RAG, select authorised sources and design evaluation questions.

01. Understand retrieval-augmented generation
02. Create an authorised source register
03. Write evaluation questions before choosing a search method

Prepare the source collection
Extract usable text, choose chunks and retain reliable metadata.

Retrieval methods
Compare embeddings, lexical retrieval, hybrid search and reranking.

Retrieve and answer within scope
Select evidence, enforce permissions and construct a grounded prompt.

Answer quality and difficult content
Handle missing evidence, verify citations and resist source instructions.

Complex questions and maintenance
Handle tables, conversational context and source updates.

Evaluation and integrated practice
Measure retrieval and answer quality, compare operating trade-offs and complete a pilot.

Assessment

Add this certificate to your resume to demonstrate your skills & increase your chances of getting noticed.

Student Ratings & Reviews

No Review Yet

AI Knowledge Assistants and Retrieval-Augmented Generation

About Course

What Will You Learn?

Course Content

Knowledge purpose and evidence
Define RAG, select authorised sources and design evaluation questions.

01. Understand retrieval-augmented generation

02. Create an authorised source register

03. Write evaluation questions before choosing a search method

Prepare the source collection
Extract usable text, choose chunks and retain reliable metadata.

04. Extract clean text without losing meaning

05. Chunk documents around useful meaning

06. Design stable chunk metadata and identity

Retrieval methods
Compare embeddings, lexical retrieval, hybrid search and reranking.

07. Understand embeddings and cosine similarity

08. Build a lexical retrieval baseline

09. Compare hybrid search and reranking

Retrieve and answer within scope
Select evidence, enforce permissions and construct a grounded prompt.

10. Choose evidence using top-k and sufficiency checks

11. Enforce access before retrieval results reach the model

12. Construct a grounded answer prompt

Answer quality and difficult content
Handle missing evidence, verify citations and resist source instructions.

13. Use structured answers and appropriate abstention

14. Verify citations at the claim level

15. Handle instructions hidden inside retrieved sources

Complex questions and maintenance
Handle tables, conversational context and source updates.

16. Answer questions involving tables and calculations

17. Resolve follow-up questions without carrying stale assumptions

18. Refresh, replace and remove source content

Evaluation and integrated practice
Measure retrieval and answer quality, compare operating trade-offs and complete a pilot.

19. Measure retrieval and answer quality separately

20. Compare latency, cost and quality trade-offs

21. Complete a knowledge-assistant pilot

Assessment

AI Knowledge Assistants and Retrieval-Augmented Generation – Final Quiz

Earn a certificate

Share this course

About Course

What Will You Learn?

Course Content

Knowledge purpose and evidence Define RAG, select authorised sources and design evaluation questions.

01. Understand retrieval-augmented generation

02. Create an authorised source register

03. Write evaluation questions before choosing a search method

Prepare the source collection Extract usable text, choose chunks and retain reliable metadata.

04. Extract clean text without losing meaning

05. Chunk documents around useful meaning

06. Design stable chunk metadata and identity

Retrieval methods Compare embeddings, lexical retrieval, hybrid search and reranking.

07. Understand embeddings and cosine similarity

08. Build a lexical retrieval baseline

09. Compare hybrid search and reranking

Retrieve and answer within scope Select evidence, enforce permissions and construct a grounded prompt.

10. Choose evidence using top-k and sufficiency checks

11. Enforce access before retrieval results reach the model

12. Construct a grounded answer prompt

Answer quality and difficult content Handle missing evidence, verify citations and resist source instructions.

13. Use structured answers and appropriate abstention

14. Verify citations at the claim level

15. Handle instructions hidden inside retrieved sources

Complex questions and maintenance Handle tables, conversational context and source updates.

16. Answer questions involving tables and calculations

17. Resolve follow-up questions without carrying stale assumptions

18. Refresh, replace and remove source content

Evaluation and integrated practice Measure retrieval and answer quality, compare operating trade-offs and complete a pilot.

19. Measure retrieval and answer quality separately

20. Compare latency, cost and quality trade-offs

21. Complete a knowledge-assistant pilot

Assessment

AI Knowledge Assistants and Retrieval-Augmented Generation – Final Quiz

Earn a certificate

Student Ratings & Reviews

Knowledge purpose and evidence
Define RAG, select authorised sources and design evaluation questions.

Prepare the source collection
Extract usable text, choose chunks and retain reliable metadata.

Retrieval methods
Compare embeddings, lexical retrieval, hybrid search and reranking.

Retrieve and answer within scope
Select evidence, enforce permissions and construct a grounded prompt.

Answer quality and difficult content
Handle missing evidence, verify citations and resist source instructions.

Complex questions and maintenance
Handle tables, conversational context and source updates.

Evaluation and integrated practice
Measure retrieval and answer quality, compare operating trade-offs and complete a pilot.