Mindrift
AI
MaterialsEngineer&PythonExpert-FreelanceAITrainer
Neural analysis suggests this role is
optimal for Mid+ candidates.
“Materials Engineer & Python Expert - Freelance AI Trainer at Mindrift. Skills: Python, Material Science, AI Training. Design computational material science problems. Write a Python reference solution”
What You'll Achieve.
Challenge a frontier AI model; Problem has an answer verifiable by code; Problem requires a specialized tool; Agent only succeeds in a small number of attempts; Task scores within range; Task quality is high; Aiming for a pass rate in the 10–30% band
Industry & Context.
Design problems that genuinely require a specialized
Participation is project-based, not permanent employment, Linux container
What They're Looking For.
Must Have
Degree in Material Science or related, 2+ years of research, applied, or teaching, Python proficiency for writing reference, Ability to design problems that genuinely require a specialized, written English (C1+)
Nice to Have
Fluency with — or willingness to independently learn — at least one scriptable package: ObsPy, instaseis, pyrocko, MITgcm, xmitgcm, flopy / MODFLOW
What You'll Do.
Design computational material science problems
Write a Python reference solution
Decide the numerical answer
Test the problem against the model
Tune the problem difficulty
Rewrite waveform scenarios
Tighten inversion parameters
How You'll Work.
Team & Collaboration
Provide feedback to ensure task quality
Communication Scope
written English (C1+)
Full Job Description
_**Please submit your CV in English and indicate your level of English proficiency.**_ Mindrift connects specialists with project-based AI opportunities for leading tech companies, focused on testing, evaluating, and improving AI systems. **Participation isproject-based, not permanent employment.** **What this opportunity involves** You design computational material science problems to challenge a frontier AI model. The problem must have an answer verifiable by code, and the problem has to require a specialized tool like ObsPy, instaseis, pyrocko, MITgcm, flopy/MODFLOW, or others. Generic data wrangling around synthesised toy data won't cut it. Each problem runs inside a sealed Linux container with the tool pre-installed and a programmatic judge that grades the model's answer. As an expert author, you: * Pick an anchor tool and design a problem that hinges on its waveform-processing kernels, geophysical inversion routines, sub-surface flow solvers, or community-validated data pipelines. * Write a Python reference solution, supply input files and model or domain definitions where needed. * Decide the numerical answer and how close the model needs to get — with a domain-appropriate tolerance — to count as right. * Test the problem against the model in batches of parallel attempts, tuning the problem difficulty until the agent only succeeds in a small number of attempts. * Once you're happy with the task, and it scores within range, the task goes to a senior reviewer in your subfield. They will provide feedback to ensure task quality is high. Calibration requires patience. You're tuning the problem against batches of parallel runs of the agent, aiming for a pass rate in the 10–30% band. Reaching that means rewriting waveform scenarios, tightening inversion parameters and solver tolerances, and watching how the agents act. You'll learn how these agents cut corners, where a simulation stalls, where a flow or inversion model converges. This time compounds in two direction
Applying for this Materials Engineer & Python Expert - Freelance AI Trainer role?
Most applicants get filtered before a human reads their resume. See if yours makes the cut.
ANONYMOUS · UNFILTERED
What do employees actually say about Mindrift?
Real rants from real employees. Read before you apply.