RETRIEVAL ENGINEERING
RAG & AI Knowledge Systems Workshop
An archived policy says ten days. The approved policy says fifteen. Find why the archive wins, repair the ranking, and prove the answer has evidence.
The local vectors are hashed word features. They do not contain model-generated embeddings.
Ranked passages
Evidence inspector
Verify an exact citation
Check whether the quote occurs in the selected passage, then decide whether it supports your claim.
Document library
Chunking microscope
Cosine similarity
Enter two illustrative numeric vectors. These values are hand-entered, not model embeddings.
Evaluation workbench
Baseline and candidate use the same documents and labeled questions. Metrics count unique documents among the retrieved chunks. The citation check requires an exact excerpt from an approved source.
Edit labeled test set
Lesson library
Worked code and walkthrough
JavaScript exercise
Input
Expected result
AI prompt studio
Review copied prompts for private source text before sharing them with an AI service.
Project library and local portfolio
Research notes
Saved revisions
Build transparent retrieval systems
Inspect exact source spans, compare ranking methods and reproduce a retrieval evaluation.
Browse interactive guides · AI learning path