Scientific Coding - Physics and Python

About this opportunity

Turing is building the SciCode project, a rigorous STEM AI training dataset focused on high-quality scientific coding tasks. As a SciCode Trainer, you will author complex physics problems, implement verified Python solutions, create discriminative test cases, perform quality checks, and refine tasks used to train and evaluate frontier AI models.

What the listing describes

• ✓Write scientific problem specifications containing one main problem and at least three logically connected sub-problems
• ✓Design problems that progressively build toward the main problem solution
• ✓Implement verified golden solutions in Python with complete unit test coverage
• ✓Design discriminative test cases that distinguish correct from incorrect model outputs
• ✓Run quality control validation checks on the Turing Central Task Platform (CTP)
• ✓Perform Tier 1 structure checks and Tier 2 quality rubric checks
• ✓Iterate on tasks based on QC feedback
• ✓Refine tasks to meet Pass@K evaluation criteria across multiple LLM judges, including GPT, Gemini, and Nemotron
• ✓Maintain high output quality and target consistent first-submission approval
• ✓Participate in review sessions, feedback meetings, and project standups during required overlap hours

Requirements shown on the source

• ✓Master's or PhD in Physics
• ✓Strong Python programming skills
• ✓Experience with scientific computing
• ✓Ability to write rigorous and well-posed scientific problems with clear constraints and expected outputs
• ✓Strong attention to detail
• ✓Ability to meet strict standards for scientific correctness, test-case discriminativeness, well-posedness, and determinism
• ✓Prior experience in AI data annotation, research, or scientific writing
• ✓Familiarity with LLM evaluation frameworks or coding benchmarks
• ✓Experience with scientific libraries such as NumPy, SciPy, SymPy, or domain-specific scientific tools
• ✓Published research or academic project experience in a STEM domain
• ✓Available for 40 hours per week
• ✓Able to provide 4 hours of overlap with PST

Relevant backgrounds mentioned

• ✓Experience creating scientific coding benchmarks
• ✓Experience evaluating AI-generated scientific or programming solutions
• ✓Experience with advanced Python-based scientific workflows