§The role
Our agents have to plan, remember, recover from failure, and not hallucinate on a live target. You'll make them better.
§What you'll do
- Post-train base models for offensive cyber tasks with RL and SFT.
- Build evals that actually predict field performance.
- Instrument long-horizon rollouts and figure out where they go wrong.
§Who we're looking for
- Background at a frontier lab, an open-source post-training project, or a serious research program.
- You can read a paper on Monday and reproduce it by Wednesday.
Ready?
Apply for AI Research Engineer
Email a resume and a paragraph on what you want to build with us.
Apply now