Physics & Python Expert - Freelance AI Trainer

Overview

This role is a project-based freelance opportunity focused on helping evaluate and improve AI systems for a leading tech company. It is not a permanent position. The work is centered on creating physics challenges that can be checked through code and used to test how well a frontier AI model performs.

What you will do

You will create computational physics problems that are designed to stretch an advanced AI model. Each task needs a verifiable solution and must depend on a specialized physics or scientific computing tool rather than only general-purpose numerical libraries. The tasks run in a locked Linux environment with the required tool already installed, and the final answer is assessed by an automated judge.

• Select a core physics tool and build a problem around its models, solvers, Monte Carlo methods, integrators, or PDE discretization features.
• Develop a Python reference solution and provide any necessary input files, initial conditions, or domain definitions.
• Determine the correct numeric result and define an appropriate tolerance that reflects the physics context.
• Run the task against the model in repeated parallel trials, then refine the setup until the success rate falls into the desired range.
• After the task is tuned and meets the required scoring band, it is reviewed by a senior specialist in the same field for quality assurance.

Calibration and task tuning

Calibration is expected to take patience and experimentation. You may need to revise boundary settings, make initial conditions more restrictive, adjust solver parameters, and observe where model attempts fail or converge. Over time, this work deepens both your understanding of the selected tool and your intuition for how advanced models handle electromagnetic, fluid, gravitational, and cosmological problems.

Requirements

Applicants should have a physics background and be comfortable with Python. Strong independent learning ability is important, especially if you need to get up to speed on a new scientific package without prior experience.

• A degree in Physics, whether theoretical, experimental, or computational, or in a closely related discipline.
• At least 2 years of experience in research, applied work, or teaching.
• Solid Python skills for creating reference implementations.
• Familiarity with, or a strong willingness to learn independently, 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 require a specialized simulation tool.
• Strong written English at C1 level or above.

Eligibility

This opportunity suits physicists or closely related specialists who are open to part-time, non-permanent project work. Even if you have not used the listed tools before, you may still be considered if you are prepared to learn independently and start contributing quickly.

How it works

The workflow is: apply, complete the qualification steps, join a project, complete the assigned tasks, and receive payment.

Working time

During active phases, the expected workload is about 10 to 20 hours per week. This is only an estimate and is not guaranteed. It applies only while a project is active.

Compensation

Contributors can earn up to the hourly equivalent of $35 on this project, with actual compensation depending on experience, contribution pace, project complexity, and scope. Pay levels may vary across different projects.

Application notes

Please provide your CV in English and include your English proficiency level with your application.