Remote Uncertainty Calibration Reviewer - Frontier AI

AI Trainer Jobs; United States; Remote; USD 28,000 - 55,000; Part time

Job summary

AuraOne is seeking a remote Uncertainty Calibration Failure Analysis Reviewer. You will evaluate model outputs against a versioned rubric, assign severity tags, and provide structured feedback to guide model retraining.

You will compare paired responses, label issues such as hallucinations or instruction deviations, and note ambiguous prompts for rubric improvements. Strong attention to detail and written reasoning are essential for consistency across long batches.

Qualifications

• Prior evaluation or annotation experience in uncertainty calibration failure analysis.
• Ability to apply multi-page rubrics consistently across long batches.
• Clear written reasoning that names the issue and rubric clause being applied.
• Reliable async availability for at least 10 hours per week.

Responsibilities

• Evaluate model outputs against a versioned rubric and assign severity tags.
• Compare paired responses and provide written rationale.
• Label hallucinations, instruction-following failures, and unsafe content with structured tags.
• Capture ambiguous prompts and route to rubric updates.

Skills

Uncertainty calibration

Annotation experience

Attention to detail

Clear reasoning

Job description

AuraOne is seeking a remote Uncertainty Calibration Failure Analysis Reviewer. You will evaluate model outputs against a versioned rubric, assign severity tags, and provide structured feedback to guide model retraining.

You will compare paired responses, label issues such as hallucinations or instruction deviations, and note ambiguous prompts for rubric improvements. Strong attention to detail and written reasoning are essential for consistency across long batches.