Mathematics Expert, AI Trainer

Fully remote, project-based work checking how well frontier AI does real mathematics. You write hard problems drawn from your own work and grade where the model's reasoning fails. $40 to $125 USD an hour, at least 10 hours a week, as an independent contractor. For people doing serious mathematics in research, industry, or applied work, in the US, UK, Ireland, Canada, or Australia.

About the role

AI models can produce a confident proof or calculation that falls apart on inspection. This role puts your experience to work finding out where. You write problems drawn from your own work, run them through frontier models, and judge the answers the way you would referee a colleague's work. Some projects involve writing code. You set your own schedule, and the work rewards careful writing as much as expertise.

What you'll do

• Build realistic tasks out of problems from your own mathematical work, along with the material needed to solve them
• Give those tasks to frontier AI models and judge the results against the standard of your profession
• Weigh paired model answers against each other and explain which one holds up
• Write grading rubrics for your tasks and set out your reasoning in writing
• Review and improve tasks written by other experts

Requirements

• Based in the United States, United Kingdom, Ireland, Canada, or Australia
• Bachelor's degree or higher, completed or in progress
• 2+ years of applied mathematics experience preferred
• Available at least 10 hours per week
• PayPal account for payment
• Willing to work without LLM assistance where the project does not permit it
• Use one email address to sign up, take the assessment, and get paid

Ideal qualifications

• Degree in mathematics, statistics, or a related field, completed or in progress
• Regular mathematical work in research, industry, or an applied setting
• Full professional or native-level written English
• Based in the United States, United Kingdom, Ireland, Canada, or Australia

Nice to have

• Programming experience, for example in Python
• Have used Claude or ChatGPT in your professional work

What success looks like

• You clear the skills assessment on your own work, with no AI assistance
• Your task proposals come from real problems in your own practice, and a frontier model genuinely fails them
• Your rubrics are specific enough that another expert would grade the same response the same way
• You act on reviewer feedback without needing to be walked through it
• You keep at least 10 hours a week going, week after week

Pay and contract

$40 to $125 USD an hour. The rate varies with your expertise, education, and experience. Minimum 10 hours a week with no weekly maximum; full-time hours are available.

Independent contractor, fully remote. Payment runs through PayPal, so you need a PayPal account. As an independent contractor, you handle your own taxes.

Getting on: a skills assessment that takes most people four to six hours. You are paid $240 for it if you pass, and you get one attempt, so take your time and do your best work. Work produced with AI tools is rejected.

After you pass: onboarding and a short training project, then you propose a task from your own practice, build it, and run it against frontier models. Each stage is reviewed before you move on, and you are paid for the work you complete.

Skills

Applied mathematics, Proof and derivation, Statistical reasoning, Modeling, Python, Rubric writing