Remote AI Evaluation Scientist — Derivations & Validation

AuraOne; United States; Remote; USD 141,000 - 190,000; Full time

Benefits offered by this job

PhD/postdoc or industry research exp.

AI tooling review experience

Multilingual fluency

Job summary

AuraOne seeks a Competitive Data Scientist for AI Evaluation (Remote) to review AI outputs, reproduce key derivations, and document the correct method so the modeling team can train on it.

You will apply multi-page rubrics, provide clear written reasoning citing methods or papers, and maintain reviewer-quality scores with at least 10 hours per week, fully remote for US-eligible contractors.

Qualifications

• Graduate-level training or equivalent applied experience in research science or a closely related field for Competitive Data Scientist — AI Evaluation (Remote) work.
• Hands-on experience publishing, teaching, or advising on the topic at a professional level.
• Comfort applying multi-page rubrics consistently across long batches.
• Clear written reasoning that cites methods, papers, or worked examples.
• Reliable async availability for at least 10 hours per week.

Responsibilities

• Review AI outputs against current research science methods, conventions, and prior work for Competitive Data Scientist — AI Evaluation (Remote) assignments.
• Reproduce or sanity-check key derivations, calculations, or experimental claims.
• Flag dimensional, methodological, and citation errors with structured severity tags.
• Capture the corrected reasoning or worked example so the modeling team can train on it.
• Adjudicate disputed answers against textbooks, papers, or community standards.
• Maintain reviewer-quality scores in inter-rater calibration cycles.

Skills

Scientific reasoning

Method validation

Citation review

Quantitative analysis

Research science

Education

Graduate-level training in research science

Job description

AuraOne seeks a Competitive Data Scientist for AI Evaluation (Remote) to review AI outputs, reproduce key derivations, and document the correct method so the modeling team can train on it.

You will apply multi-page rubrics, provide clear written reasoning citing methods or papers, and maintain reviewer-quality scores with at least 10 hours per week, fully remote for US-eligible contractors.