Veridex
Trust layer for people and AI agents built on World ID and Worldchain, with portable reputation and crypto-backed accountability.
What I built
Built a trust protocol that ties real humans and AI agents back to a portable, verified reputation layer. Veridex uses World ID for proof-of-personhood, aggregates reputation signals from sources like GitHub and supporting credentials, and exposes that trust through onboarding, dashboards, contracts, reviews, and a public verification flow.
Challenges
The hardest part was making the identity and crypto layers feel integral instead of bolted on. World ID session handling, QR flow, and app connectivity added real implementation complexity, and the project also needed to balance deterministic scoring, AI-supported evidence interpretation, and agent accountability in one coherent product under hackathon time pressure.
Architecture
The system pairs a Next.js 14 frontend with an Express and TypeScript backend, Supabase for persistence and auth, Gemini for evidence extraction and grounded assistance, and Worldchain for ETH staking and reputation incentives. The scoring pipeline stays deterministic and auditable rather than depending on an opaque LLM loop, while agent credentials let third-party services verify the human behind an AI agent.
Results
Veridex finished as a top-ten team at Catapult 2026 and became one of my strongest web3 projects because the blockchain layer was materially meaningful. Worldchain staking created transparent skin in the game, World ID anchored the system in verified personhood, and the overall product showed how crypto, AI agents, and trust infrastructure can work together in a real application.