Bevisly

A platform where skill claims come with structured, verifiable proof — built full-stack with Supabase, PostgreSQL, and a row-level security model designed for multi-role data isolation.

Skills are claimed everywhere and verified nowhere. The product problem was making endorsements mean something. The technical problem was designing multi-role data isolation at the database layer — so the security model is a constraint, not a client-side hope.

Supabase for RLS-based multi-role auth so security lives in the database, not the frontend. React and TypeScript on the front so the data model surfaces cleanly in the component tree. Next.js for SSR and SEO.

Owned the product end-to-end: defined the two-role model (candidate and employer), chose database-level row security over application-level checks, and designed those policies before writing UI, then shipped 8+ AI features on top of a security layer most side projects skip entirely. Designed the Employer Responsibility Score and Candidate Reliability Score to make ghosting visible and costly — an incentive-design decision, not a feature request.