All AI training jobs

Mercor · STEM & research

PhD Physicist — Frontier Reasoning Trajectory Writer

Listed on Mercor as “PhD Physicist - Frontier Reasoning Trajectory Writer

$80-$120/hrRemoteContractPaid in USD

What this actually is

You design problems that stump current AI models, evaluate AI reasoning against the correct answer, write rubrics, and provide expert feedback. Often the highest-paid category because the expertise pool is small. The platform title (PhD Physicist - Frontier Reasoning Trajectory Writer) reflects the rate band and the expertise required, not the day-to-day work.

Can you do this on your visa?

F-2 / F-4 / F-5 / F-6: open. E-1 to E-7: needs concurrent-employment permit. D-2 / D-4 students: S-3 permit, 20 hr/week cap. D-10 / D-8: case by case.

Korean tax on USD income

First 5 years in Korea: foreign-source income only taxed if remitted into Korea. After year 5: worldwide income. Full tax guide.

Original posting from Mercor

PhD Physicist - Frontier Reasoning Trajectory Writer

What you'll do

Produce "golden trajectories" for hard physics problems: complete, step-by-step solution walkthroughs where every reasoning step is sound, the final answer is correct, and the work is organized around a sensible upfront plan. You'll work in Studio, iterate with a frontier LLM to draft and refine the reasoning, and produce trajectories that are clean enough to use as training data. The problems are provably hard - we screen them by running a frontier LLM on each problem five times and requiring it to fail at least once.

Domain focus

We need physicists with a theoretical or computational specialization in one of:

  • Astrophysics (theory / numerical)
  • Condensed matter (theory)
  • Quantum physics / quantum information
  • Particle / high-energy physics

We are not looking for experimentalists, software engineers, or data scientists without a physics PhD background.

Requirements

  • PhD in physics, or in the final 1-2 years of a PhD program, from a credible research institution
  • Daily-driver fluency in LaTeX (for derivations) and Python (for numerical sanity checks and plotting)
  • Comfort with iterative LLM workflows - willing to nudge models with hints, identify reasoning errors, and shape an exploration into a clean trajectory
  • Detail-oriented; comfortable producing publication-grade derivations

Time commitment

Around 15 hours per week, roughly one to two tasks per week. Each task averages about 7 hours of writer time.

How to apply

Submit your application and complete the short technical assessment. Strong applicants will be invited to a follow-up interview.

Quoted from Mercor’s public listing on 2026-06-02. We don’t edit platform copy; honest framing is in the title and the “what this actually is” block above.