AI startup · Europe · 2026 → now · Full stack
AI agent evaluation
Companies putting AI agents in front of their customers without knowing how they can fail. Full stack development of a platform that tests AI agents automatically and reports every failure.
Context
A European startup builds tools so companies can trust the AI agents they put in front of their customers.
What I built
I work on the full stack development of its evaluation platform: the interface, the API and the data model. The platform tests an agent automatically and groups what it finds by type of risk: made-up answers, attempts to get it to break its rules, personal data it shouldn’t reveal, and straying from what the company allows.
Every failure is recorded so both the engineering and compliance teams can review it.
Built with
TypeScript, React and Node.
Outcome
A platform that looks for an agent’s failures before its users find them. It’s my main work today.
Technologies
- TypeScript
- React
- Node