Materials Engineer & Python Expert - Freelance AI Trainer

About Mindrift

Mindrift facilitates project-based AI opportunities for skilled professionals with leading technology firms. Our focus is on the rigorous testing, evaluation, and enhancement of artificial intelligence systems. Please note that participation is project-specific and does not constitute permanent employment.

Project Involvement

In this role, you will be responsible for creating computational material science challenges designed to test a cutting-edge AI model. Each problem must have a solution that can be verified programmatically and necessitate the use of specialized tools such as ObsPy, instaseis, pyrocko, MITgcm, flopy/MODFLOW, or similar. Standard data manipulation tasks with synthetic data will not suffice. Problems will be executed within isolated Linux containers equipped with the necessary pre-installed tools and a programmatic judge to assess the AI model's responses.

Expert Author Responsibilities

• Select a core tool and devise a problem that relies on its specific functionalities, such as waveform processing, geophysical inversion, sub-surface flow simulation, or community-validated data pipelines.
• Develop a Python-based reference solution, provide necessary input files, and define model or domain parameters.
• Determine the correct numerical answer and establish an acceptable tolerance range for the AI model's output, appropriate to the specific domain.
• Iteratively test the problem against the AI model using parallel execution batches, adjusting the difficulty to achieve a success rate within a target range (typically 10-30%).
• Submit the finalized task for review by a senior expert in the relevant subfield to ensure high quality and accuracy.

Calibration and Learning

This process requires careful calibration and patience. You will fine-tune problem parameters by running the AI model in parallel batches, aiming for a success rate between 10% and 30%. This involves refining waveform scenarios, adjusting inversion parameters and solver tolerances, and observing the AI's behavior. Through this, you will gain insights into how AI models handle complex seismic, oceanographic, and sub-surface flow problems, and deepen your expertise with the chosen specialized tool.

What We Are Looking For

This freelance opportunity is ideal for material scientists and engineers with Python experience who are interested in part-time, project-based work. Ideal candidates will possess:

• A degree in Material Science or a related discipline.
• A minimum of two years of experience in research, practical application, or teaching.
• Proficiency in Python for developing reference solutions.
• Familiarity with, or a strong commitment to independently learning, at least one of the following scripting packages: ObsPy, instaseis, pyrocko, MITgcm, xmitgcm, flopy / MODFLOW, or GeoPandas.
• The ability to design problems that genuinely require specialized solvers.
• Excellent written English skills (equivalent to C1 level or higher).

Even if you don't have prior experience with the specific tools listed, you are encouraged to apply if you are prepared to learn them independently and quickly.

Project Workflow

The process involves applying, successfully completing qualification assessments, joining a project, executing assigned tasks, and receiving payment.

Estimated Project Workload

During active project phases, tasks are estimated to require approximately 10 to 20 hours per week. This is an estimate and the actual workload may vary based on project demands.

Compensation Details

Contributors on this project can earn up to the equivalent of $35 per hour, with the final rate dependent on the individual's expertise and performance. Compensation can differ across projects based on their scope, complexity, and the specific skills required.

Additional Information

Please submit your CV in English and clearly indicate your English proficiency level.