About the Role
We're a small, ambitious team building the next generation of AI training data — grounded in verifiable truth rather than noisy human annotation or web-scraped text. Our core thesis: frontier model capability breakthroughs are gated on data breakthroughs. We integrate formal systems — physics simulators, scientific databases, formal proof systems, executable tests, and oracle databases — to produce training data that is correct by construction.
We're hiring both Research Scientists and Research Engineers to join us on the ground floor. You'll work directly with the founding team on some of the hardest open problems in AI data generation, with outsized impact on the direction of the company.
What You'll Do
- Design and build systems that generate high-quality, verifier-grounded training data for large language models, robotics, and scientific AI applications.
- Develop pipelines that leverage formal systems (physics simulators, proof systems, executable tests, oracle databases) to validate and produce correct-by-construction data at scale.
- Run experiments to measure data quality and its downstream impact on model performance.
- Collaborate closely with the founding team to define research directions and prioritize high-leverage bets.
- Translate research insights into robust, production-quality implementations (Research Engineers) or push the frontier of what's possible with novel methods (Research Scientists).
What We're Looking For
We're open to a wide range of experience levels (0–12 years). We value depth of thinking and a track record of building or discovering things that work.
- Strong background in machine learning, deep learning, or a related field — either through industry experience, a graduate degree, or demonstrated self-directed work.
- Experience with large language models (LLMs), generative models, or reinforcement learning is a strong plus.
- Comfort working in at least one of: robotics, scientific computing, formal methods, or data pipeline engineering.
- Solid software engineering skills; ability to prototype quickly and iterate (Python proficiency expected).
- Research Engineers: emphasis on scalable system design, data infrastructure, and turning research ideas into reliable pipelines.
- Research Scientists: emphasis on novel method design, experimentation rigor, and publishing or equivalent demonstrated research output.
- Genuine curiosity about how data shapes model capability — and opinions about what the field is getting wrong.
Dealbreaker: Must be legally authorized to work in the United States without visa sponsorship. Visa sponsorship is not available for this role.
Compensation & Benefits
- Salary: $100,000 – $300,000 USD annually (range reflects both scientist and engineer tracks across experience levels)
- Early-stage equity
- High-autonomy, high-ownership environment at a seed-stage company with strong investor backing
Location
- On-site — San Francisco, CA
- This is a fully in-person role; remote work is not available.