Scientific Computing SME – Math & Python at Turing

Scientific Computing Subject Matter Expert | Contractor | Worldwide Remote

Turing is seeking highly specialized Subject Matter Experts (SMEs) in Mathematics, Physics, Chemistry, or Biology to contribute to a cutting-edge scientific benchmark development project. You'll design research-grade computational science problems used to evaluate frontier AI systems — specifically targeting Gemini 3 Pro — following the rigorous SciCode benchmark format.

Based in San Francisco, Turing is the world's leading research accelerator for frontier AI labs, partnering with global enterprises to deploy advanced, reliable AI systems at scale.

What You'll Do

• Design computationally intensive, research-level scientific problems across domains such as Numerical Linear Algebra, Quantum Mechanics, Reaction Kinetics, or Computational Genomics.
• Write complete, verified Python reference solutions using NumPy, SciPy, and related scientific libraries.
• Develop comprehensive pytest test suites validating correctness, numerical accuracy, convergence, and algorithmic stability.
• Structure each problem as a main problem with 2–4 supporting subproblems, scaffolded bottom-up to build toward the full solution.
• Deliver all work as structured Jupyter notebooks (.ipynb) with clearly organized Prompt, Background, Solution, and Testing cells.
• Ensure diversity across topics and subdomains — no repetition of similar algorithmic patterns.

Requirements

• PhD in Physics, Chemistry, Mathematics, Biology, or a related computational field (preferred); Master's degree with 5+ years of relevant research or industry experience may be considered.
• Strong proficiency in Python for scientific computing.
• Hands-on experience with NumPy, SciPy, SymPy, and Pandas.
• Deep understanding of numerical methods, algorithms, and computational complexity.
• Ability to design problems with precise, verifiable numerical outcomes.
• Experience writing automated tests with pytest.
• Familiarity with numerical stability, floating-point arithmetic, and error analysis.

Problem Quality Standards

• Problems must be computationally intensive — iterative solvers, numerical simulations, matrix decompositions, or adaptive methods required. Simple closed-form calculations are not acceptable.
• Solutions must be standalone Python functions, not scripts.
• Each deliverable must be a complete Jupyter notebook covering the main problem and all subproblems.

This is a high-impact opportunity to shape how frontier AI systems are evaluated on real-world scientific reasoning and computation.