Particle and Nuclear Physicist for AI Benchmark Design

Design graduate-level computational physics challenges that evaluate advanced AI systems on realistic research workflows, from running simulations and interpreting results to designing experiments and extracting hidden information from data.

You will create original problems grounded in particle and nuclear physics workflows, then test them with advanced AI models and refine them to reach an appropriate level of difficulty. Problems may require exact answers from fully defined, multi-step setups or strategic planning of queries and experiments to infer information that is not directly observable.

• Create computational problems requiring skilled use of specialized scientific software.
• Develop research-level tasks involving simulations, result interpretation, experiment design, and data-driven discovery.
• Design problems that test strategic measurement choices, interpretation of partial results, and efficient narrowing of possible explanations.
• Test each problem against advanced AI models and iteratively improve its difficulty and quality.
• Write Python-based problem setups, oracle functions, and solution validators.

• Graduate-level training in a relevant STEM field, with an MS, PhD, or equivalent research experience preferred.
• Deep hands-on experience with particle and nuclear physics software, including scikit-hep and related high-energy physics Python tools.
• Experience with particle physics data analysis, cross-section calculations, renormalization group calculations, and perturbative QCD.
• Proven proficiency with at least one relevant scientific software library through research publications, open-source contributions, or professional work.
• Strong Python skills and the ability to work independently while incorporating feedback into problem designs.
• Comfort working in Linux or terminal environments and remote compute sandboxes.
• Ability to identify practical tool limitations and edge cases, and to create challenges that reward careful reasoning rather than surface-level pattern matching.

• Experience with Monte Carlo event generation or collider phenomenology.
• Experience with other specialized software used in this domain.
• Experience across multiple relevant scientific domains or tools.
• Familiarity with benchmark or evaluation design.
• A background in scientific teaching or designing exams and problem sets.
• Experience with computational reproducibility and containerized environments.

• Remote, hourly engagement.
• Availability of at least 15 to 20 hours per week is required.

Hourly compensation ranges from $70 to $100.