Computational Mathematics AI Evaluation Expert

OpenTrain AI, Inc.; United States; Remote; USD 90,000 - 120,000; Full time

Job summary

OpenTrain AI, Inc. seeks a remote contract contributor to craft advanced computational mathematics tasks for AI training. You will design multi-step problems, assemble datasets, and validate results across Python, R, Julia, and C/C++ in Linux environments.

Responsibilities include building reference solutions, defining grading criteria, and ensuring numerical accuracy, stability, and reproducibility. Strong mathematical background and independent work are essential.

Qualifications

• PhD or postdoctoral experience in mathematics, statistics, or related field.
• Strong programming ability in Python, R, Julia, C/C++, Bash.
• Experience in Linux/terminal environments and numerical computing.
• Ability to design, implement, test, and validate computational algorithms.

Responsibilities

• Create realistic, multi-step computational tasks in self-contained environments.
• Prepare datasets, equations, model definitions, constraints, and outputs.
• Build reference solutions and computational workflows across languages.
• Set grading criteria for numerical accuracy, convergence, and stability.
• Define tolerances, stopping criteria, and reproducibility controls.
• Investigate floating-point issues, conditioning, and solver failures.
• Document formulations, outputs, limitations, and validation methods.

Skills

Python

R

Julia

C/C++

Bash

Linux

Numerical methods

Mathematical reasoning

Education

PhD or postdoc

Tools

NumPy

SciPy

SymPy

pandas

scikit-learn

JAX

PyTorch

CVXPY

Docker

Conda

Job description

The work

You will create realistic, multi-step computational mathematics tasks in self-contained terminal environments. These tasks will evaluate AI agents working through research and engineering problems.

You will combine mathematical formulation, implementation, debugging, automated validation, and careful review of computational results.

• Create problems involving numerical analysis, optimization, statistics, probability, mathematical modeling, differential equations, dynamical systems, operations research, and computational geometry.
• Prepare datasets, equations, model definitions, constraints, initial conditions, expected outputs, and documented assumptions.
• Build reference solutions and computational workflows using Python, R, Julia, C/C++, Bash, or another suitable language.
• Set grading criteria for numerical accuracy, convergence, complexity, feasibility, stability, and mathematical correctness.
• Define tolerances, stopping criteria, reproducibility controls, automated tests, edge cases, and valid alternative implementations.
• Investigate floating-point precision, solver failures, conditioning, convergence, and performance issues.
• Document mathematical formulations, expected outputs, known limitations, and validation methods.

What it pays and takes

This is a remote contract role for an individual contributor focused on specialized AI training task design, mathematical evaluation, and computational validation. The listing does not specify a pay rate.

The role is listed as entry level, but the requirements call for advanced technical experience and independent work in computational mathematics.

• Time requirement: 20+ hours per week.
• Location: Remote and open worldwide.
• Language: English.
• Education or experience: A Ph.D., postdoctoral experience, or equivalent advanced technical experience in mathematics, statistics, or a closely related field.
• Programming: Strong ability in Python, R, Julia, C/C++, Bash, or a comparable language.
• Environment: Experience working in Linux or terminal-based environments.
• Technical background: Numerical methods, optimization, statistics, mathematical modeling, or scientific computation.
• Core abilities: Independently implement, test, and validate computational algorithms; understand numerical stability, error analysis, mathematical assumptions, and reproducibility.
• Helpful tools and experience: NumPy, SciPy, SymPy, pandas, scikit-learn, JAX, PyTorch, CVXPY, optimization solvers, probabilistic programming, differential-equation libraries, numerical linear algebra, Monte Carlo methods, Bayesian inference, graph algorithms, computational geometry, Docker, Conda,
• Additional helpful experience: Mathematical programming challenges, benchmark tasks, automated graders, AI coding evaluations, research software, HPC workflows, or performance-critical numerical code.

About AI training work

AI training work uses human-created examples, evaluations, and feedback to improve how artificial intelligence systems perform. Specialists are needed for difficult work such as checking mathematical reasoning, testing code, and judging whether computational results are accurate and reliable.