AI Benchmark Problem Designer: Computational Statistics

Obsidian; Seattle (WA); On-site; USD 120,000 - 180,000; Part time

Job summary

Obsidian is seeking a Computational Statistics and Applied Mathematics Expert to design graduate-level problems for AI benchmarks. You will craft tasks requiring real scientific software and multi-step workflows that test planning, data interpretation, and experimental design.

You will test problems against state-of-the-art AI models, iterating until the target difficulty is met. This role supports remote work and part-time hours (15–20 per week), with Seattle area preference.

Qualifications

• Graduate-level training in statistics, applied mathematics, or a closely related field.
• Proven proficiency with at least one specialized statistical, mathematical, or scientific software package, demonstrated through research publications, open-source contributions, or professional work
• Strong Python skills — you'll be writing problem setups, oracle functions, and solution validators
• Ability to work independently and refine problem designs based on feedback
• Comfortable working in a Linux/terminal environment with remote compute sandboxes
• Available for at least 15–20 hours per week

Responsibilities

• Create problems requiring skilled use of specialized statistical, mathematical, or scientific software packages.
• Test problems against state-of-the-art AI models and refine them until the target difficulty is reached.
• Design and interpret numerical results from fully defined setups.
• Plan experiments and queries to uncover information not directly visible.
• Work through testing loops against modern AI models and iterate accordingly.

Skills

Python programming

Statistical analysis

Applied mathematics

Problem design

Independent work

Education

MS or PhD in statistics / applied math

Tools

R

Python

Matlab/Scilab

Linux/Unix

Job description

Obsidian is seeking a Computational Statistics and Applied Mathematics Expert to design graduate-level problems for AI benchmarks. You will craft tasks requiring real scientific software and multi-step workflows that test planning, data interpretation, and experimental design.

You will test problems against state-of-the-art AI models, iterating until the target difficulty is met. This role supports remote work and part-time hours (15–20 per week), with Seattle area preference.