What the work actually is

Each task you produce is a small, self-contained physics problem set with code attached. You write a main problem plus at least three sub-problems that build toward it logically — not three unrelated exercises bolted together. You then implement the golden solution in Python with full unit test coverage, and design test cases that are discriminative: a model that hand-waves the physics or swaps a sign should fail them, while a genuinely correct alternative implementation should pass. Determinism matters, so anything relying on unseeded randomness, floating-point luck, or loose tolerances will come back to you.

Submissions go through Turing's Central Task Platform. Tier 1 checks structure; Tier 2 applies quality rubrics for well-posedness, scientific correctness, and test discriminativeness. Tasks are then run against multiple LLM judges (GPT, Gemini, Nemotron) under Pass@K criteria, and you iterate on whatever the QC feedback flags. Throughput matters, but rework rate matters more — the stated target is consistent L1 approval on first submission, which means the discipline is front-loaded into how carefully you specify the problem before you write a line of code.

What the screen looks for

  • A real graduate physics background (Master's or PhD) that you can talk about concretely: subfield, methods, what you computed and how.
  • Python for scientific work, specifically NumPy/SciPy/SymPy fluency and the habit of writing tests rather than eyeballing plots.
  • Judgment about problem design — whether you can spot an ambiguous prompt, an under-constrained boundary condition, or a test that any half-correct solution would pass.
  • Willingness to work inside someone else's rubric and revise against QC notes without relitigating them.

Logistics

Remote, contractor engagement with no paid or medical leave, quoted as 40 hrs/week for eight weeks. The hard scheduling constraint is a four-hour daily overlap with PST for standups, reviews, and feedback sessions — the rest is asynchronous authoring. Hiring is restricted to Bangladesh, Brazil, Colombia, Egypt, Ghana, India, Indonesia, Kenya, Nigeria, Pakistan, Turkey, and Vietnam. Pay is undisclosed on the listing; ask for the per-task or hourly rate and the rework policy before you commit.