What the work actually involves

You receive sets of AI-generated responses — often two or more competing answers to the same prompt — and rank or score them against a rubric supplied by the customer. Dimensions typically include instruction adherence, factual accuracy, tone appropriateness, and genuine usefulness to the stated audience. The score is only half the deliverable: each judgment requires a written rationale that a reviewer who never saw the prompt could follow and agree with. Expect to flag the failure modes that rubrics are built to catch — responses that read beautifully but rest on nothing, responses that are exhaustive yet unusable, quiet internal contradictions, and confident factual errors.

What the screen is looking for

micro1's screening is AI-led and conversational. It probes whether you have genuinely assessed written work against criteria — marking exam scripts, moderating grades, supervising theses, reviewing submissions, designing assessment instruments — and whether you can articulate why a piece of writing fails, not just that it does. The strongest signal is calibration language: how you handle borderline cases, how you keep your scoring consistent across a long batch, how you separate surface fluency from substance. No AI or machine learning background is required. Prior RLHF or annotation work is a plus, never a gate.

Logistics

  • Fully remote, contractor engagement, asynchronous — work is claimed from a queue rather than scheduled into shifts.
  • Observed pay band of $90–140/hr, varying by project and assessed depth; rates are what has been reported, not a guarantee.
  • Designed to sit alongside teaching, examining, or other professional commitments, though projects carry real turnaround deadlines.
  • Collaboration happens through a content review platform plus written threads with project managers and senior reviewers.
  • Preference for candidates with a completed degree from an accredited institution in the US, Canada, UK, Ireland, Australia, or New Zealand, and first-language or fully equivalent written English.