About the job
Physical Superintelligence Fellowship
Join the Physical Superintelligence PBC (PSI) as we strive to create cutting-edge AI systems that revolutionize physics discovery. Originating from prestigious institutions such as Harvard, MIT, Johns Hopkins, Oxford, the Institute for Advanced Study, and the Perimeter Institute, we are assembling an outstanding team of physicists, AI researchers, and engineers.
At PSI, we are committed to transforming scientific discovery through AI, recognizing the unique challenges physics presents that require profound domain expertise to address effectively. We are dedicated to developing innovative systems to tackle these challenges.
Fellowship Overview
The PSI Fellowship offers an exceptional opportunity for physicists eager to engage at the forefront of AI-driven scientific discovery. We are looking for talented researchers who can integrate rigorous physics thinking with modern AI technologies, possessing both deep domain expertise and practical experience in AI system development.
Fellows will collaborate closely with PSI’s founding team on pivotal research challenges, aiming to make significant contributions to our technical roadmap and, where suitable, contribute to publishable research.
Program Duration: 3–6 months, full-time, based on project scope and mutual compatibility.
What to Expect
Collaborate directly with PSI leadership and technical advisors on impactful research projects.
Flexibility to engage in projects that align with your expertise, ranging from benchmark development to agent architectures to physics evaluation frameworks.
Access to advanced computational resources and tools for AI-physics research.
Potential pathway to full-time employment for outstanding performers (though not guaranteed).
Research Areas
Fellows will have the opportunity to work on diverse projects, including:
AI-for-Physics Benchmarks: Crafting robust evaluations to determine if AI systems can genuinely reason about physics as opposed to merely pattern-matching solutions.
Physics Discovery Agents: Designing and assessing agentic systems capable of formulating hypotheses, planning experiments, and interpreting results.
Verification & Interpretability: Developing frameworks to confirm the validity of AI-generated physics insights and to comprehend how models represent physical knowledge.
Post-Training for Scientific Reasoning: Enhancing foundational models’ reasoning capabilities about physics through techniques such as Reinforcement Learning from Human Feedback (RLHF).

