Remote Japanese AI Evaluation Specialist

AuraOne, Inc.; United States; A distancia; USD 55.000 - 83.000; Jornada completa

Ventajas ofrecidas por este puesto de trabajo

Remote work

Descripción de la vacante

AuraOne, Inc. is seeking an independent contractor to remotely review japanese generalist evaluation prompts and responses, applying a versioned rubric to assign severity tags. You will compare paired outputs, provide rationales, and flag issues like hallucinations or unsafe content.

The role emphasizes careful annotation, durability across large batches, and willingness to calibrate against gold-standard examples on a weekly cadence.

Formación

• Prior evaluation, annotation, or human-rater experience on japanese generalist evaluation or adjacent content.
• Clear written reasoning that names the issue and the rubric clause being applied.
• Reliable async availability for at least 10 hours per week.

Responsabilidades

• Evaluate japanese generalist evaluation model outputs against a versioned rubric and assign severity tags for Japanese ChatGPT User assignments.
• Compare paired responses and pick the stronger answer with a written rationale.
• Label hallucinations, instruction-following failures, and unsafe content with structured tags.
• Capture ambiguous prompts and route them back to the program team for rubric updates.
• Maintain reviewer-quality scores by calibrating against gold-standard examples each week.
• Document recurring failure modes so the modeling team can target them in the next training run.

Conocimientos

Model output evaluation

Rubric-based annotation

Severity tagging

Inter-rater calibration

Japanese generalist evaluation

Eligibility

Descripción del empleo

AuraOne, Inc. is seeking an independent contractor to remotely review japanese generalist evaluation prompts and responses, applying a versioned rubric to assign severity tags. You will compare paired outputs, provide rationales, and flag issues like hallucinations or unsafe content.

The role emphasizes careful annotation, durability across large batches, and willingness to calibrate against gold-standard examples on a weekly cadence.