Materials Engineer & Python Expert - Freelance AI Trainer

Overview

This project-based opportunity is for specialists who want to help improve advanced AI systems for major technology companies. The work is not a permanent role; instead, it involves short-term project participation focused on evaluating and strengthening model performance.

Role Summary

As a domain expert, you will create computational materials science problems that are difficult enough to challenge a frontier AI model. Each task must have a solution that can be checked programmatically, and it must rely on a specialized tool such as ObsPy, instaseis, pyrocko, MITgcm, flopy/MODFLOW, GeoPandas, xmitgcm, or a similar scientific package. Simple toy-data manipulation is not sufficient. Tasks are executed in a locked Linux environment with the required tool already installed, and a programmatic judge evaluates the answer.

What You Will Do

You will be responsible for building tasks that test real scientific reasoning and tool usage, then refining them until they meet the required difficulty band.

• Select a core tool and create a problem that depends on its waveform processing, inversion methods, subsurface flow solvers, or validated geoscience data workflows.
• Develop a Python reference solution and provide any necessary input files or domain/model definitions.
• Define the correct numerical outcome and set an appropriate tolerance for judging how close a model’s answer must be.
• Run the task repeatedly against parallel model attempts and adjust the challenge until the pass rate falls within the target range.
• Submit the task for senior review in your subfield once it performs within the expected difficulty range.

Calibration and Difficulty Tuning

This work requires patience and iteration. You will adjust waveform scenarios, tighten inversion settings, and tune solver tolerances while observing how the model behaves. The goal is to achieve a pass rate of roughly 10% to 30% during testing. Over time, you will gain deeper expertise in the anchor tool and build practical intuition for how frontier models handle seismic, oceanographic, and subsurface flow challenges.

How the Process Works

The workflow follows a simple sequence: apply, complete qualification steps, join a project, finish assigned tasks, and receive payment.

Time Commitment

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

Compensation

Contributors may earn up to the equivalent of $35 per hour on this project, depending on experience level and speed of delivery. Pay can differ from one project to another based on scope, complexity, and the expertise required.

Application Notes

Please submit your CV in English and include your English proficiency level.

Additional Information

Participation is project-based and non-permanent. Candidates without prior exposure to the listed tools may still apply, provided they are prepared to learn independently and start quickly.