Materials Engineer & Python Expert - Freelance AI Trainer

Overview

Mindrift brings together domain specialists for project-based AI work with leading technology companies. The focus is on creating, checking, and refining AI systems through short-term assignments rather than ongoing employment.

Role summary

In this assignment, you will build computational materials science problems that can be used to evaluate a frontier AI model. Each challenge must have a code-verifiable answer and should depend on a specialized tool such as ObsPy, instaseis, pyrocko, MITgcm, flopy/MODFLOW, GeoPandas, xmitgcm, or a similar scriptable package. Simple toy-data exercises or generic data cleanup tasks are not suitable. The problems are executed in an isolated Linux container with the required tool already installed, and a programmatic judge scores the model’s response.

What you will do

As a contributor, you will select a core tool and design a task that relies on its waveform-processing capabilities, geophysical inversion methods, subsurface flow solvers, or other validated domain pipelines. You will also create a Python reference solution, prepare input data and any model or domain definitions needed, determine the correct numerical result, and define a realistic tolerance for acceptance.

You will then test the task in batches of parallel attempts, adjusting the setup until the model succeeds only occasionally. Once the task reaches the expected scoring range, it is reviewed by a senior specialist in the same field to verify quality and provide feedback.

Calibration and task tuning

This work requires patience and careful iteration. The goal is to tune each challenge so that the pass rate lands roughly in the 10% to 30% range. That often means revising waveform scenarios, tightening inversion settings and solver tolerances, and observing where models fail or take shortcuts. Over time, you gain deeper knowledge of the anchor tool and a practical understanding of how advanced models handle complex seismic, oceanographic, and subsurface flow problems.

Requirements

We are looking for material scientists or engineers with Python experience who are open to part-time, non-permanent project work. The ideal contributor has a degree in materials science or a closely related field, at least 2 years of research, applied, or teaching experience, and strong English communication skills at C1 level or higher.

You should be comfortable writing Python reference solutions and either already know, or be ready to independently learn, at least one scriptable package from the list provided. You also need the ability to create problems that truly depend on a specialized solver. Prior experience with the named tools is helpful but not required, as long as you are prepared to ramp up quickly and work independently.

Working arrangement

Applicants must send a CV in English and include their English proficiency level. The work is project-based and not a permanent role. Estimated effort during active phases is around 10 to 20 hours per week, depending on project needs. This estimate applies only while a project is active and is not a guaranteed workload.

Compensation

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

Process

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