Member of Technical Staff (Frontier Robotics Evals)

Your day-to-day will look like:

• Designing robotics benchmarks.
• Running evaluations of frontier AI models.
• Working with actual robots (e.g. bimanual arms, humanoids, quadrupeds).
• Developing our open-source evaluation harness used both internally and externally.
• Engineering the low-latency inference infrastructure for robotics foundation models.
• Measuring recursive self-improvement of AI in robotics capabilities.
• Building towards the METR plot for physical automation.
• Architecting our research roadmap on prioritizing task families, model families, and embodiments.
• Scaling to hundreds of parallel real-world robotics evals within months.

• In person at our office in San Francisco
• Exceptional software engineering skills with agentic tools (e.g. Claude Code, Codex)
• Exceptional communication skills
• Willingness to learn and move fast

Not required, but a plus:

• Experience in AI evaluations (e.g. Inspect AI/published benchmarks)
• Experience in robotics (e.g. Master’s/PhD/robotics lab experience)
• Context on AI safety and the future trajectory of AI (e.g. BlueDot fellowship, ARENA, SPAR, MARS, PIBBSS, Pivotal, LASR Labs, CBAI, ERA, Anthropic Fellows Program, MATS, Astra)

Aside from technical skills, we find that strong candidates have these personal qualities:

• Reasoning transparency. Arriving at conclusions through verbalized/written reasoning instead of relying on unexamined hunches.
• Overcommunication over undercommunication.
• Prioritizing what is important. Research directions follow a power law on their impact.
• Gives honest feedback to teammates and implements honest feedback from teammates.
• Strong team player.

• Annual salary of $170k–$300k.
• For very experienced and exceptional candidates, we are open to exploring paying much higher than this stated range.
• Generous equity as an early employee.

• Comprehensive healthcare insurance (medical, dental, vision).
• Unlimited PTO.
• Unlimited sick leave to prioritize your well-being.
• All work travel paid, covering flights, hotels, and expenses.
• Team retreats twice a year.

Please apply through the application form. We give rolling offers and prioritize candidates who can start as soon as possible.

Refer a candidate. $5k for successful referrals.

Robocurve evaluates frontier AI robotics systems and informs the public about physical automation (i.e. “METR/Epoch for robots”). We develop benchmarks and open-source tooling (e.g. Inspect Robots) to evaluate AI on robots. Robocurve is a YC-backed Public Benefit Corporation. We work closely with frontier AI labs of all sizes and academic groups.