Mathematician for AI Model Evaluation

Apply advanced mathematical reasoning and computational problem solving to improve and evaluate large language models. You will design rigorous math problems, produce clear, logically structured solutions, verify numerical results with code, and review model outputs to identify errors and missing arguments. This role blends formal mathematical thinking, practical Python implementation, and the use of formal theorem proving to push the reasoning capabilities of state of the art models.

• Accelerate frontier AI research by contributing high quality data and evaluation material, and by working with advanced training pipelines and expert researchers who specialize in coding, reasoning, STEM, multilinguality, multimodality, and agents.
• Help enterprises move AI from proof of concept to reliable, proprietary intelligence by producing evaluation artifacts and methodologies that measure real world performance and deliver measurable impact on business outcomes.

• Create original, challenging mathematics problems that probe multi step reasoning, abstraction, and proof skills of large language models.
• Solve problems independently and write detailed, logically ordered solutions with clear justifications and steps.
• Review and annotate model generated solutions, identify mathematical mistakes or omitted arguments, and provide precise corrections and feedback.
• Help define new evaluation benchmarks based on mathematics curricula spanning early undergraduate through PhD level topics.
• Design exact, closed ended prompts and produce reliable Python solutions for computational tasks using approved scientific libraries, verifying numerical answers.
• Work on theorem prover tasks using Lean, translating problems and proofs into formal language and confirming formal proofs compile correctly.
• Contribute to improving your own workflows by learning to leverage AI tools to be a more effective analyst.

• Strong foundation in mathematics at the level expected for engineering entrance exams and for graduate or PhD level programs.
• Proven ability to break down complex mathematical concepts into clear, simple explanations, using visuals or examples where helpful.
• Experience writing reliable Python code for numerical or symbolic computation, and familiarity with scientific libraries.
• Experience or willingness to work with the Lean theorem prover for formalization and verification of proofs.
• Excellent structured written communication, ability to provide constructive feedback and detailed annotations.
• Creative and lateral thinking, strong research and analytical skills, and the ability to work independently in a remote setting.
• Technical requirements: access to a desktop or laptop with a stable internet connection.

• Pursuing or holding a Master’s, PhD, or Postdoctoral degree in Mathematics, Applied Mathematics, Statistics, or a closely related field. Candidates in these programs are eligible and encouraged to apply.
• Demonstrated ability to analyze and solve complex math problems using structured, logical approaches.
• Facility explaining math concepts in plain language, with examples and visuals when appropriate.

• Remote contractor assignment, freelance engagement. This role is contractor status only, there is no medical or paid leave provided.
• Time commitment options: 20 hours per week, 30 hours per week, or 40 hours per week.
• Minimum required commitment: at least 4 hours per day and a minimum of 20 hours per week, with 4 hours of overlap with Pacific Standard Time, to enable collaboration.
• Potential for contract extension based on performance and project needs.

Compensation details are not specified in the listing.

• Applicants pursuing or holding graduate level degrees in mathematics related fields are eligible and encouraged to apply. No additional work authorization details were provided in the source.

Candidates who meet the qualifications, especially those enrolled in or holding Master’s, PhD, or Postdoctoral degrees in relevant fields, are encouraged to apply. The listing does not provide further application steps or platform specific instructions.