I build audit-first AI and engineering prototypes for materials science, scientific software, and risk-aware decision workflows.

Current focus:

• HEA / MPEA candidate audit interfaces
• descriptor-level review workflows
• claim-boundary-safe AI demos
• Molecule Identity Audit: a fail-closed molecular identity, metadata consistency, and claim-boundary audit showcase.
• engineering sandbox prototypes with explicit non-production boundaries
• local-first software experiments with reproducible audit trails

These repositories are research and software prototypes. They do not claim experimental validation, production readiness, safety certification, or autonomous material recommendation.

A static MVP baseline for passive high-entropy alloy candidate auditing.

It focuses on descriptor-level review, evidence separation, claim boundaries, and human-review-required audit snapshots.

A local sandbox prototype for digital-twin-style engineering risk routing.

It separates modeled thermal drift from low-temperature brittleness proxy behavior and records simulated routing outputs with audit-traceable boundaries.

Audit before action.
Evidence before claims.
Boundaries before automation.