The work
You will adapt genuine computational work from your own research into self-contained evaluation tasks for AI models. Day to day, that means defining a main physics problem and connected subproblems, implementing a Python reference solution, writing tests, specifying evaluation metrics, and revising the task after structured feedback. You may also review other researchers' submissions against a short rubric.
The relevant research areas are condensed-matter and statistical physics; atomic, molecular, and optical physics; photonics and electromagnetics; astrophysics, cosmology, and gravitation; nuclear, particle, and plasma physics; fluid dynamics and nonlinear physics; and continuum mechanics. The platform screens for direct ownership of computational research, not general familiarity with coding or machine learning. Candidates should be ready to explain the physical model, numerical method, debugging history, validation evidence, and assumptions behind a specific project.
Tasks require a Python reference solution. For prior research code, Python is preferred, but the core requirement is that you have written your own code to simulate or analyze physical systems. Strong submissions derive their difficulty from the physics rather than missing context, and include tests that distinguish physically correct solutions from superficially plausible outputs.
Contract and schedule
Pay band: $100/hr
The work is independent-contractor, task-based, and has no fixed weekly schedule. The platform estimates about two hours per task, although review and revision needs can vary, and there is no stated cap on task volume. The workflow appears suited to independent, asynchronous completion, but the listing does not explicitly confirm an asynchronous or remote-work policy. Accepted experts undergo identity and background verification.