Astrophysicist - AI Trainer

The newest models can vibe-code a whole feature in minutes. Someone has to check whether that code actually works, and that someone is you. You'll run agentic coding sessions on real engineering tasks, analyze what the models produce, and red-team them until they fail in interesting ways. Every failure you document teaches the next generation of coding agents. What you’ll actually do Vibe-code with frontier models on real tasks: fix a bug, extend an API, ship a feature, and see how far the model gets on its own. Analyze the AI's work against production standards: find the bugs, explain the failure, and rate the model's reasoning. Red-team the models to expose unsafe code, faked test passes, and confidently wrong solutions before real users hit them.