Your mission is to advance practical AI safety, evaluation, and governance methods that can survive real production conditions.
What you will own
- Design research questions tied to consequential AI-system failure modes.
- Build reproducible evaluations, datasets, and experimental methods.
- Translate findings into runtime controls, evidence, and product requirements.
- Publish clear internal research records and challenge weak assumptions.
- Work across research, engineering, policy, and red-team functions.
What we look for
- Deep experience in machine learning, AI safety, evaluation, or an adjacent technical field.
- Strong experimental design and statistical reasoning.
- Ability to implement and validate research systems, not only describe them.
- Excellent technical writing and reproducibility habits.
- Comfort working on uncertain, adversarial, and high-consequence problems.
How we hire
No resume is required at this stage. We begin with structured questions so your judgment and actual work can carry the application. We do not use AI to rank, score, summarise, or reject candidates.