Job Description
Job DescriptionAbout the Role
We're a newly founded, Y Combinator–backed startup at the frontier of AI training data. Our core thesis: model capability breakthroughs require data breakthroughs. We build "correct by construction" training data for cutting-edge AI models — grounding every data point against formal verification systems rather than relying on slow human annotation or noisy web-scraped text.
We're hiring both Research Scientists and Research Engineers as early team members. You'll work directly with the founders to shape the direction of the company and tackle fundamental problems in AI data quality, large language models, robotics, and AI for science.
What You'll Do
-
Design and build pipelines for generating verifier-grounded training data at scale, using physics simulators, formal proof systems, scientific databases, executable tests, and oracle databases.
-
Develop methods to project fundamental laws of physics, biological facts, and self-consistent logic into natural language data for improved AI reasoning.
-
Conduct original research to advance the state of the art in data quality, data generation, and model training for LLMs, robotics, and scientific AI.
-
Collaborate closely with frontier AI labs and internal teams to understand data needs and translate them into scalable solutions.
-
Publish findings, prototype novel approaches, and iterate rapidly in a small, high-ownership environment.
What We're Looking For
Research Scientist profile:
-
Strong background in machine learning, deep learning, or a related field (PhD or equivalent research experience preferred).
-
Track record of original research contributions — publications, patents, or demonstrable project impact.
-
Experience with LLMs, generative models, robotics learning, or AI for science.
-
Comfort working with formal systems, symbolic reasoning, or verification methods is a strong plus.
Research Engineer profile:
-
Strong software engineering fundamentals with hands-on experience building ML systems or data pipelines at scale.
-
Proficiency in Python and modern ML frameworks (e.g. PyTorch, JAX).
-
Experience with simulation environments, physics engines, or formal verification tooling is highly desirable.
-
Ability to translate research ideas into reliable, production-quality systems.
For both roles:
-
Genuine intellectual curiosity and drive to work on hard, open-ended problems.
-
Comfort with ambiguity and a strong bias toward action in an early-stage environment.
-
Must be legally eligible to work in the United States without visa sponsorship.
Compensation & Benefits
-
Salary: $100,000 – $300,000 USD annually (range reflects both Scientist and Engineer levels and candidate experience)
-
Early-stage equity
-
Benefits package commensurate with a well-funded early-stage startup
Location
This is a full-time, on-site role based in San Francisco, CA. We believe in the value of in-person collaboration, especially at this early stage. Visa sponsorship is not available.