Founding Researcher

Join an early-stage AI safety research team working to develop AI systems that follow legal requirements. You will design and run research experiments across evaluations, training, and runtime controls, owning projects from conception through publication.

• Design and run experiments across research on testing AI agents against legal requirements, training them to follow those requirements, and controlling them at runtime.
• Build evaluation environments, training pipelines, and runtime harnesses.
• Work with legal experts and academic advisors to ground research in legal practice.
• Publish research and release benchmarks and tools.

• At least one year of research or machine learning engineering experience in industry, academia, or independent projects, with a research-oriented background and published work, preprints, or equivalent.
• A bachelor's, master's, or doctoral degree in computer science, mathematics, machine learning, or a related field, or equivalent practical experience.
• Experience in AI safety, alignment, or evaluations, demonstrated through research, open-source projects, or industry work.
• Experience with evaluations, reinforcement learning environments, or post-training, including developing evaluation rubrics and benchmarking models across specific capabilities.
• Strong Python engineering skills and familiarity with the modern machine learning stack, including PyTorch.
• Comfort navigating open-ended research problems without predefined solutions.
• Commitment to AI safety and developing systems that follow legal requirements.
• Must be authorized to work in the United States without visa sponsorship.

Annual base salary: $185,000 to $250,000. Visa sponsorship is not available.

On-site in San Francisco, California, United States.