The work
Help evaluate how AI models handle difficult problems in formal methods, computational science, or a relevant scientific field. The focus is on research, technical judgment, and clear written feedback, not production engineering.
- Design realistic technical scenarios and problem sets in your area of expertise.
- Write expert-level reference solutions and grading rubrics.
- Review AI-generated answers for correctness, depth, and sound use of subject knowledge.
- Explain your technical judgments in clear, structured feedback.
Requirements and work details
- Education: A PhD or doctorate in a relevant scientific or computational field.
- Experience: At least 10 years in your field, with current or recent hands-on work.
- Expertise: Formal methods, computational science, mathematics, physics, chemistry, or biology.
- Communication: Strong written English for technical explanations and structured feedback.
- Work style: Able to work independently, asynchronously, and remotely.
- Time: 20+ hours per week.
- Location: Remote, but open only to candidates in eligible countries; the role is not worldwide.
- Employment type: Part-time contractor.
- Pay: Not specified in the role details.
Helpful experience
Research publications, open-source contributions, recognized work in your field, or experience reviewing technical work are useful.
- Relevant review work may include peer review, code review, grading, or editing.
- Specialist experience may include formal methods or theorem proving with Lean or Coq, computational biology or genomics, or condensed-matter or quantum physics.
How to apply
OpenTrain is a platform where people find AI training and data-labeling roles and begin applications. Apply on OpenTrain with your resume, then complete the application on the hiring site.
About AI training work
AI training is human work that helps AI models learn from examples and feedback. This role uses expert-written problems and reviews of model answers, where deep subject knowledge helps judge whether the reasoning is sound.