Work Type
Remote
Location
Global
Compensation
$300 per approved task
About this opportunity
Overview
We are seeking Earth Sciences experts to create realistic terminal-based scientific tasks for Terminal Bench Science. The work focuses on building computational workflows that AI agents can execute in a terminal environment, with an emphasis on objectively verifiable scientific outputs.
This is a contractor assignment with an expected duration of 5 weeks and a planned start date next week. The role requires at least 6 hours per day and a minimum of 40 hours per week, with 4 hours of overlap with PST.
What You'll Do
- Translate authentic Earth-science workflows into self-contained terminal benchmark tasks.
- Prepare geospatial, climate, atmospheric, geological, hydrological, or oceanographic datasets.
- Build reproducible computational environments using scientific libraries and command-line tools.
- Create expert solutions using Python, R, Bash, Julia, or domain-specific software.
- Design tasks involving geospatial analysis, time-series processing, numerical modeling, interpolation, forecasting, remote sensing, or environmental risk analysis.
- Define objective grading criteria for scientific outputs, model behavior, data transformations, and spatial or temporal accuracy.
- Validate coordinate systems, units, timestamps, missing-data handling, and scientific assumptions.
- Create automated tests for numerical tolerances, file formats, metadata, and reproducibility.
- Debug issues involving geospatial projections, large datasets, dependencies, performance, and numerical stability.
- Document input data provenance, expected outputs, edge cases, and limitations.
Requirements
- Ph.D., postdoctoral experience, or equivalent advanced technical experience in Earth Sciences or a closely related field.
- Strong expertise in at least one area such as climate science, atmospheric science, geophysics, oceanography, geology, hydrology, remote sensing, environmental modeling, or Earth-system science.
- Strong programming skills in Python, R, Julia, Bash, or another scientific programming language.
- Hands-on experience with scientific data processing, numerical modeling, geospatial analysis, environmental datasets, or time-series analysis.
- Comfort working independently in Linux or terminal-based environments.
- Ability to build, debug, and validate reproducible scientific computational workflows.
- Strong understanding of scientific quality control, spatial and temporal data, uncertainty, and numerical accuracy.
Preferred Qualifications
- Experience with NumPy, pandas, SciPy, xarray, rasterio, GeoPandas, Cartopy, GDAL, or similar scientific and geospatial tools.
- Experience with scientific data formats such as NetCDF, HDF5, GeoTIFF, shapefiles, or GRIB.
- Experience working with climate, weather, satellite, seismic, oceanographic, geological, or hydrological datasets.
- Familiarity with Docker, Conda, Git, CI/CD, automated testing, or HPC environments.
- Research software engineering, scientific benchmarking, or automated grader development experience.
- Experience evaluating AI coding or terminal agents, or developing tasks and evaluations for AI systems.
- Publications or open-source contributions in Earth, environmental, geospatial, or computational sciences.
Contract Details
- Commitment: at least 6 hours per day and 40 hours per week.
- Schedule overlap: 4 hours with PST.
- Employment type: contractor assignment with no medical or paid leave.
- Duration: 5 weeks.
- Payment: $300 per approved task.
