SuperAnnotate AI, IncNorth America

Physics Experts - PhD required - (Freelance, Remote, 20+ openings)

Project-Based

Description

Role: Engineering

Rate: $10.00–$85.00 USD/hr

Skills: Data Annotation, Physics

Project Description

This project involves reviewing Al-generated responses, evaluating the reasoning quality

and step-by-step problem-solving, and providing expert feedback that helps models

produce answers that are accurate, logical, and clearly explained. You will assess solutions

for accuracy, clarity, and adherence to the prompt; identify errors in calculations,

methodology, or conceptual understanding; fact-check physics information; write

high-quality physics explanations and model solutions that demonstrate correct methods;

and rate and compare multiple Al responses based on physics correctness and reasoning

quality.

Compensation

Compensation is competitive and varies based on education level, institution, and prior experience.

Hiring Profile

  • • You hold a PhD in Physics.
  • • You have developed or critically reviewed complex math content (e.g., problem
  • banks, proofs, textbook sections, research notes).
  • • Published in a peer-reviewed journal.
  • • English proficiency C1 and above with excellent writing skills.
  • • Bonus: Data labeling, RLHF, or Al model evaluation experience

Next steps

  • Apply via partner site - Braintrust will send you an email with detailed instructions.
  • Qualification exam (1–2 hours): evaluates your role-specific expertise. Compensation provided upon passing.
  • Project exam (1–2 hours): evaluates your adherence to project guidelines and ability to produce math-specific deliverables. Compensation provided upon passing.

Commitment

  • We expect a minimum of 17–20 hours per week, however considering the project scope and timelines the preferred and recommended commitment would be 8 hours/day.

#LI-MS1 #LI-Remote

Skills

PhysicsData Annotation

Want AI to find more roles like this?

Upload your CV once. Get matched to relevant assignments automatically.

Try personalized matching