Physical Sciences Expert – Terminal Bench

Role overview

We are seeking physical sciences experts to develop realistic, terminal-based scientific tasks for Terminal Bench Science. You will translate authentic workflows in physics, chemistry, materials science, astronomy, and computational science into reproducible benchmark environments.

The role involves creating scientific inputs, computational models, executable solutions, automated tests, and objective grading criteria. You will help evaluate whether AI agents can reason through scientific problems, use command-line tools, debug calculations, and produce reliable scientific artifacts.

Open positions: 120

Core domains

Physics, chemistry, materials science, computational physics, computational chemistry, astronomy and cosmology, thermodynamics, quantum mechanics, statistical mechanics, and materials modeling.

Responsibilities

• Design multi-step terminal tasks based on realistic physical-science workflows.
• Develop self-contained computational environments with pinned dependencies and scientific software.
• Create input datasets, molecular structures, simulation parameters, experimental data, or model configurations.
• Implement expert solutions using Python, Bash, C/C++, Julia, or domain-specific tools.
• Develop automated tests for numerical accuracy, physical consistency, convergence, and output structure.
• Create tasks involving simulations, numerical modeling, data fitting, optimization, spectroscopy, molecular analysis, or scientific visualization.
• Define numerical tolerances, units, boundary conditions, and expected scientific behavior.
• Validate that tasks are reproducible and run successfully without runtime downloads.
• Debug dependency, precision, solver stability, performance, and file-format issues.
• Communicate scientific assumptions and computational limitations clearly to reviewers.

Requirements

• Ph.D., postdoctoral experience, or equivalent advanced technical experience in a relevant physical-science discipline.
• Strong programming skills in Python, C/C++, Julia, Bash, or another scientific programming language.
• Experience working in Linux or terminal-based environments.
• Experience with numerical methods, scientific modeling, simulations, or quantitative data analysis.
• Ability to create and validate computational scientific workflows independently.
• Strong understanding of units, numerical precision, physical constraints, and scientific reproducibility.

Nice to have

• Experience with NumPy, SciPy, pandas, matplotlib, SymPy, JAX, or similar libraries.
• Familiarity with molecular dynamics, quantum chemistry, finite-difference methods, Monte Carlo methods, optimization, or statistical modeling.
• Experience with OpenMM, ASE, RDKit, Psi4, LAMMPS, GROMACS, or other domain-specific software.
• Familiarity with Docker, Conda, Git, CI systems, and automated testing.
• Experience with HPC systems or performance-sensitive scientific workloads.
• Experience evaluating AI coding or terminal agents, research software engineering, benchmark-task or automated-grader development, computational-science publications or open-source contributions, or converting experimental and research workflows into reproducible packages.

Engagement details

• Contractor assignment; no medical benefits or paid leave.
• Full-time commitment of 40 hours per week, with four hours of PST overlap per day.
• Contract duration: five weeks; immediate start.
• Payment: $300 per approved task. Average task handling time is 6–8 hours.
• Partner fee: $175.
• Listed locations: Bangladesh, India, Pakistan, Indonesia, Egypt, Ghana, Nigeria, Turkey, Brazil, and Colombia.

Selection process

1. Complete the Interest Check Form (ICF), using the same email address submitted on the agency portal.
2. Complete the mandatory 20-minute Terminal Bench screening test.
3. Complete one 40-minute Physical Sciences assessment in astronomy and cosmology, physics, materials science, or chemistry.
4. Delivery manager review.