Mathematician for Scientific AI Evaluation

Design original, executable scientific computing challenges that help evaluate the limits of advanced AI systems. This role focuses on creating rigorous mathematics problems with a coding component, grounded in research, real-world data, open-source projects, or scenarios you develop. Key Responsibilities

• Author original research problems in scientific computing that are difficult for current frontier AI models to solve.
• Source problem material from published papers, Kaggle datasets, open-source repositories, or self-designed scenarios.
• Create scientific prompts and the grading criteria used to determine correct answers.
• Test and calibrate tasks against frontier models, with tasks released only when strong models fail more often than they succeed.

• PhD in mathematics, applied mathematics, computational mathematics, or a closely related field.
• Demonstrated depth, with a coding focus, in at least two of these areas: numerical linear algebra, computational mechanics, and computational finance.
• Working proficiency in Python for scientific computing.
• Comfort using Git and GitHub, running code in Docker, and contributing through a pull-request workflow with automated quality checks.

• Publications in peer-reviewed journals.
• Scientific software development or research engineering experience.

• Remote, hourly independent contractor engagement.
• Six-week project with an immediate start date.
• Part-time commitment of at least 20 hours per week.

• $70 per hour.

• Submit a resume and completed application form.
• Participate in a 25-minute conversational interview covering your background, experience, and motivations.
• Receive follow-up within a few days regarding next steps and onboarding.