Astrophysics AI Benchmark Designer (Computational)

Obsidian; San Francisco (CA); On-site; USD 55,000 - 83,000; Full time

Job summary

Obsidian is seeking a Computational Astrophysics & Cosmology Expert to design graduate-level computational problems that use real scientific software for research-grade work—running simulations, interpreting results, and designing experiments. You will craft benchmarks that probe strategic reasoning, test them against state-of-the-art AI models, and refine problems until the target difficulty is achieved.

A MS/PhD in astrophysics or cosmology with hands-on Python and Astropy experience is

Qualifications

• Graduate-level training in a relevant STEM field (MS, PhD, or equivalent)
• Proven proficiency with at least one listed scientific software library as shown through publications or work
• Strong Python skills for problem setups, validators, and oracle functions
• Ability to work independently and refine problem designs based on feedback
• Comfortable in a Linux/terminal environment with remote compute sandboxes
• Available for at least 15–20 hours per week

Responsibilities

• Design graduate-level computational problems using real scientific software
• Create problem setups, oracle functions, and solution validators
• Test problems against state-of-the-art AI models and refine for target difficulty
• Plan a sequence of queries or experiments to uncover hidden information
• Iterate designs based on feedback and model performance
• Collaborate on benchmarks and reproducibility in containerized environments

Skills

Python

Linux/Terminal

Independent work

Problem design

Education

Graduate degree in STEM

Tools

Astropy

Containerization

Job description

Obsidian is seeking a Computational Astrophysics & Cosmology Expert to design graduate-level computational problems that use real scientific software for research-grade work—running simulations, interpreting results, and designing experiments. You will craft benchmarks that probe strategic reasoning, test them against state-of-the-art AI models, and refine problems until the target difficulty is achieved.

A MS/PhD in astrophysics or cosmology with hands-on Python and Astropy experience is