Project Meteor — AI Response & Rubric Evaluation — OneForma

Last verified: 30 July 2026 · Source: oneforma.com/jobs/meteor/

What is this project?

You're reviewing and improving AI-generated "golden responses" — the reference answers used to train and grade AI models — along with the evaluation rubrics used to score them. The specific goal called out in the listing: catching and correcting "critical (P0)" errors, i.e. the most serious class of mistake.

What would I actually do day to day?

• Review AI-generated reference answers for quality and accuracy.
• Evaluate and improve the rubrics used to grade those answers.
• Identify and fix critical (P0) issues.
• Work across both open-ended and close-ended evaluation tasks, following detailed guidelines.

Requirements

• Native-level fluency in one of the project languages.
• Strong written communication skills and excellent attention to detail.
• Basic familiarity with medical, financial, and legal topics — the listing calls this out specifically, suggesting some content spans regulated/technical subject matter.
• Reliable internet connection.

Is there an exam?

Not described on this listing. OneForma's general certification process likely applies, but this page doesn't restate it.

Pay

Not publicly disclosed. Listed only as "Fixed Rate Per Hour" with no number shown on the public page.

Who should apply

Detail-oriented native speakers who enjoy analytical work — reading carefully, catching subtle errors, and improving how quality is measured, rather than producing original content themselves.

How to prepare

TTA Academy's Prompt Engineering Masterclass overlaps well with this role — understanding what makes a prompt/response pair good or bad is core to both (TTA Academy's own recommendation).

Part of the OneForma vendor guide. TTA Academy is not affiliated with OneForma.