The work
You will create realistic computational physical-science tasks for AI models and define what a correct solution must include. You will also assess whether generated technical work is accurate, reproducible, and physically sound.
The work combines physical-science expertise with Python and computational research. You will give clear, technically grounded judgments on model outputs.
- Create task instructions that test calculations, code, simulations, models, and technical analysis.
- Set up reproducible environments, reference solutions, evaluation criteria, rubrics, and graders.
- Review AI-generated Python code, calculations, simulations, models, and other technical deliverables.
- Find invalid assumptions, numerical errors, physically impossible results, incorrect methods, and reproducibility failures.
- Set objective standards for correct solutions and explain your technical judgments.
What it pays and takes
This is expert-level contract work for people with prior paid experience training, evaluating, benchmarking, or improving AI systems on scientific or technical tasks.
- Pay: $60 to $120 per hour equivalent, paid per task that meets the project specifications.
- Time: Less than 20 hours per week.
- Work type: Part-time contract work.
- Location: Open to candidates in the listed countries across the Americas, Europe, Asia, Africa, and the Middle East, including the United States, Canada, the United Kingdom, India, Japan, and Singapore.
- Language: Professional fluency in English.
- Experience: Expert-level experience with AI training or model evaluation involving scientific or technical content.
- Python: Strong practical proficiency with Python.
- Science: Deep expertise in at least one physical-science specialization.
- Research: Hands-on experience with computational modeling, simulation, numerical analysis, data analysis, or reproducible research.
- Evidence: Public technical work such as computational publications, preprints, public simulation or analysis code, research profiles, open-source contributions, reproducible notebooks, models, datasets, benchmarks, or experiments.
Helpful background
Experience creating AI-training tasks, reference answers, rubrics, graders, or benchmark problems is useful. Familiarity with scientific tools such as NumPy, SciPy, SymPy, Astropy, PyTorch, JAX, or comparable libraries is also valuable.
Relevant specialties include physics, astronomy and cosmology, materials science, chemistry, computational physics, computational chemistry, and scientific simulation. Experience validating simulations or models against analytical results, observations, or experimental data is helpful.
How it works
You apply here in about two minutes and finish on the hiring platform. The next step is a short AI interview (technical roles include a coding challenge). Passing it is how you qualify for paid tasks, so take it soon after you are invited.
The hiring platform's application for this job asks for your resume, your LinkedIn profile, your phone number, your location and answers to 8 screening questions.
About AI training work
AI training is the human work behind systems that learn from examples, including reviewing code, rating model answers, and checking technical outputs. OpenTrain is a free platform where people find AI training opportunities, build a profile, and apply through the listed process.