OpenAI for Science aims to harness AI to accelerate the process of scientific research by building models and an AI-powered platform that speeds up discovery and helps researchers everywhere do more, faster.
Requirements
- Regularly use frontier models in their own research
- Either know or are eager to learn modern AI and run AI experiments end-to-end
- Open-source contributions to mathematical science or AI tooling
- Experience building or curating domain datasets and benchmarks
- Have used ChatGPT to do calculations and prove or improve lemmas in your field of study
Responsibilities
- Assist in designing and building frontier AI models that are great at solving frontier mathematical sciences problems
- Build high-quality scientific datasets and synthetic data pipelines (symbolic, numeric, and simulator-based)
- Design reinforcement and grading signals for physics and run reinforcement learning/optimization loops to improve model reasoning
- Define and run evals for scientific reasoning, derivations, simulations, and literature grounding; track progress over time
- Shape training and evaluation
- Guide how to wire models to scientific tools
- Work with the academic community to speed up adoption and impact
Other
- Hold a current or recent academic position in mathematical sciences (mathematics, theoretical physics, theoretical computer science) or a related field
- Move easily between theory and code, and are eager to contribute technically as well as academically
- Are strong scientific communicators
- Care about rigor and reproducibility in scientific results
- Communicate clearly to both scientists and AI engineers; you like collaborating across teams and with academia
