AI Trainer (Earth & Environmental Sciences)

You will create demanding computational tasks for AI systems and assess whether their technical answers are correct, reproducible, and scientifically sound. Your reviews may cover code, geospatial analyses, simulations, forecasts, and other Earth and environmental science outputs.

You will define clear instructions, reference solutions, and evaluation criteria. You will also check results against observations, measurements, or established models and identify issues with coordinate systems, spatial or temporal data, assumptions, uncertainty, numerical methods, and reproducibility.

• Create realistic computational Earth and environmental science tasks.
• Write task instructions, reproducible environments, reference solutions, and evaluation criteria.
• Review model-generated Python code, geospatial analyses, simulations, forecasts, and technical deliverables.
• Set objective standards for correct technical solutions and assess technical submissions.
• Validate computational results against observations, measurements, or established models.

This is a remote, part-time contractor opportunity for candidates with intermediate experience. The schedule is flexible, with a commitment of at least 15 hours per week and a workload of less than 20 hours per week.

• $60 to $120 per hour equivalent, paid per task that meets the project specifications.
• Professional fluency in English.
• Prior paid experience training, evaluating, benchmarking, or improving AI systems on scientific or technical tasks.
• Strong practical proficiency in Python.
• Deep expertise in at least one Earth or environmental science specialization.
• Hands-on experience with geospatial analysis, modeling, simulation, forecasting, remote sensing, statistics, or reproducible research.
• Relevant areas include geoscience, geology, geophysics, atmosphere and climate, meteorology, ocean and marine science, hydrology, remote sensing, environmental science, and sustainability.
• Useful tools include NumPy, SciPy, pandas, xarray, GeoPandas, GDAL, rasterio, and NetCDF, or comparable software.
• Experience creating AI-training tasks, reference answers, rubrics, graders, or benchmark problems is useful.
• Publications, preprints, open-source scientific software, datasets, models, notebooks, maps, simulations, or research profiles can demonstrate relevant work.