Physics & Python Expert - Freelance AI Trainer

Overview

This is a project-based opportunity with a team that pairs subject-matter specialists with AI training work for major technology companies. The work centers on evaluating, testing, and improving AI systems, and it is not a permanent role.

What you will do

Your focus will be on creating computational physics challenges that can be checked programmatically and used to stress-test a frontier AI model.

• Select a suitable simulation or scripting tool and build a problem around its underlying physics, numerical methods, Monte Carlo routines, or PDE solvers.
• Prepare a Python-based reference implementation and provide any necessary input data, initial conditions, or domain setup files.
• Define the expected result and set an appropriate discipline-specific tolerance for evaluating correctness.
• Run repeated test batches against the model, then refine the task difficulty until success rates sit in the target range.
• Submit well-calibrated tasks for senior review once the exercise meets the quality and performance target.

Task calibration and workflow

Calibration is intentionally iterative and may take patience. The goal is to tune each challenge so that the model succeeds only in a limited share of parallel attempts, typically around 10% to 30%. Achieving that may require adjusting field setups, tightening initial conditions, and refining solver parameters while observing how the model behaves.

Through this process, you build stronger command of the tool you choose and gain practical insight into how advanced models handle difficult problems across electromagnetics, fluids, gravity, and cosmology.

Who this suits

This role is intended for physics specialists with Python experience who are comfortable working on part-time, non-permanent projects.

Eligibility and expectations

• A degree in Physics, whether theoretical, experimental, or computational, or in a closely related field.
• At least 2 years of research, applied, or teaching experience.
• Working knowledge of Python for building reference solutions.
• Either hands-on fluency with, or a strong independent learning mindset for, at least one scriptable physics package such as FEniCS / DOLFINx, OpenFOAM, Meep, MPB, openEMS, Geant4, PYTHIA8, ROOT / PyROOT, WarpX, REBOUND, MESA, CAMB, CLASS, or bilby.
• The ability to create problems that truly depend on specialized simulation software rather than generic numerical libraries alone.
• Strong written English at C1 level or above.
• Applicants without prior experience in the listed tools are still encouraged to apply if they are ready to learn quickly and contribute independently.
• Submit your CV in English and clearly state your English proficiency level.

How the process works

You apply, complete the qualification steps, join a project, work on tasks, and then receive payment for completed work.

Time commitment

During active project periods, the expected workload is about 10 to 20 hours per week. This is only an estimate and is not a guaranteed number of hours.

Compensation

Contributors may earn up to the equivalent of USD 35 per hour, with actual pay varying by level and contribution speed. Compensation can differ from one project to another depending on scope, complexity, and the expertise required.