What the work actually involves

You will spend most of your time reading model output closely and deciding whether it is good English — not just correct English. A typical task queue mixes several formats: rate two candidate summaries against a source text and justify the ranking; correct a model's grammar-correction attempt when the model over-corrects a dialect form; rewrite a stilted paragraph so it reads naturally without changing its meaning; or tag a passage for tense inconsistency, dangling modifiers, ambiguous reference, or register mismatch. Alongside ratings you write short prose rationales, and those rationales are the deliverable that matters most — they are what the lab reads when deciding how to change the model.

The judgment calls are subtler than proofreading. You will regularly be asked to distinguish an actual error from an acceptable regional or stylistic variant, to say whether a hedge is appropriate caution or evasion, and to decide when a fluent sentence is nonetheless wrong because it misrepresents the source. Rubrics are supplied and revised mid-project; part of the job is flagging where a rubric fails to cover a case rather than silently guessing.

What the screen looks for

Mercor's screening is AI-led: a recorded conversational interview after a résumé and profile review. It probes for concrete professional history with English text — house styles you have worked to, annotation schemes you have applied, volumes and turnaround times you have handled — and for whether you can name linguistic phenomena precisely rather than gesturing at "awkward phrasing." Expect follow-up questions that push on the reasoning behind a judgment you just gave. Interest in AI is welcome but not a requirement; demonstrable editorial or linguistic depth is.

Logistics

  • Fully remote, asynchronous; you pull tasks rather than sit in meetings.
  • Hours are flexible and generally uncommitted week to week — some contributors work a few hours, others closer to part-time, depending on queue volume.
  • Pay band observed at approximately $50/hr; project-dependent and not guaranteed.
  • Work is project-based and can taper or expand as the lab's needs shift.