Math Expert AI Evaluation Fellowship Remote

What You'll Do8

• Design rigorous math questions and prompts that test an AI model's reasoning.
• Review AI-generated responses for mathematical accuracy and step-by-step logic.
• Flag errors in proofs, calculations, and formal arguments.
• Write clear feedback that explains where a response fails or succeeds.
• Assess whether a model's solution meets the standards of advanced mathematical proof.
• Align questions with advanced math domains such as algebra, analysis, topology, or logic.
• Work on a self-directed schedule across active projects.
• Apply the same evaluation standards across many model responses.

Requirements8

• Enrolled in a PhD program or hold a doctorate in mathematics or a related field.
• Postdoctoral researchers qualify.
• Strong command of advanced mathematical reasoning and proof-writing.
• Background in formal logic.
• Ability to judge AI-generated content for accuracy and logical consistency.
• No prior AI or technical experience required.
• Comfort with remote, project-based contract work and flexible hours.
• F-1 students may qualify through CPT or OPT; STEM OPT is not supported.

Who Should Apply

The ideal candidate is a math PhD student, postdoc, or doctorate holder who enjoys writing proofs and finding subtle gaps in logical arguments. You can shape a 5 to 20 hour week around research, teaching, or industry work because the schedule is asynchronous and project-based. People who need guaranteed full-time hours or a permanent academic appointment will not find that here. A common reason applicants score low is weak evidence of advanced proof-writing or formal logic training, even when they have strong computational math skills. Work authorization also matters: F-1 students may qualify through CPT or OPT, but STEM OPT is not supported, and a school CPT course requirement may not match the program.

Salary Insight