Mathematics QA / AI Trainer (3+ yrs Math + Python) — Contract, Remote, Part-Time

Job Description

OpenTrain partners with leading AI companies to build safer, smarter systems. We hire subject-matter experts to audit and improve model reasoning. We’re seeking Mathematics QA / AI Trainers to review math prompts and model responses, verify step-by-step reasoning, and author improved solutions and prompts that raise quality and rigor.

What you’ll do

Audit math Q&A items and model outputs for correctness, completeness, notation, and structure.

Evaluate responses against detailed QA rubrics; assign ratings and write concise rationales.

Draft improved prompts, reference solutions, and error-aware feedback (SFT / RLHF / evaluation).

Run quick validations in Python (functions / sanity checks; NumPy / SymPy / Pandas helpful).

Flag edge cases, propose rubric refinements, and help maintain dataset consistency across topics.

Minimum Qualifications

3+ years in mathematics, applied math, or a quantitative role (e.g., research, analytics, teaching).

Bachelor’s in Math or related field ( Master’s / PhD preferred ).

Hands-on Python for quick checks / validation.

Able to evaluate algebra, calculus, linear algebra, probability, and statistics reasoning.

Meticulous with mathematical notation, proofs, and solution structure.

Clear written English; comfortable following strict QA rubrics and writing constructive feedback.

Reliable computer / internet; able to pass an AI interview and QA calibration.

Contract & Schedule

Type : Contract (remote).

Hours : Less than 20 hours / week; minimum 15 hours / week (flexible).

Duration : 3–6 months (potential extension).

Compensation : USD $8–$55 / hour (target $40 / hour ).

Estimated scope : ~40 hours for initial engagement, with potential for additional work.

About OpenTrain

OpenTrain equips cutting-edge AI with high-quality, human-grounded feedback. As a contractor, you’ll help shape how next-generation AI reasons through university-level mathematics.

Location

Global (remote).