Requirements ↔ regulation mapping
automotive manufacturing delivered
Context
An automotive manufacturer must continuously map internal requirements to the regulations that govern them. The mapping is document-heavy and judgment-heavy: thousands of lines on both sides, and a wrong or missed correspondence is a compliance problem.
Constraint
The engineers who own the requirements had to stay in control of every accepted mapping. The system could propose; it could not decide. Existing material lived in spreadsheets and had to flow in without a re-keying project.
What we built
A mapping system that ingests requirement and regulation documents, proposes correspondences with AI assistance, and routes every proposal through a human review workflow — accepted, rejected or corrected, with the reasoning captured.
What it changed
Mapping became reviewable work instead of manual cross-referencing: the reviewers' judgment moved to where it matters — deciding — while the machine handled the finding and pairing. The review trail doubles as compliance evidence.
Stack
Commodity components: Python (FastAPI), React.