Build working systems
Demonstrate real workflows, evidence and failure modes rather than only architecture diagrams.
ABOUT INSURMIND
A serious side project about a difficult boundary.
InsurMind explores where language models help insurance underwriting—and where they should stop. The work focuses on source evidence, deterministic rules, organisational knowledge and human authority.
Preserve uncertainty
Keep missing, ambiguous and contradictory source facts visible to the reviewer.
Measure the controls
Test rule grounding, omitted-clause detection, retrieval quality and decision stability.
COLLABORATE