What the work actually involves

You write technical prompts in Spanish at the level of your own research — the kind of question a graduate student or working scientist in your subfield would actually ask — and then judge how the model answers. Some tasks are straight accuracy grading: is the mechanism right, is the reagent stoichiometry plausible, does the protocol reflect how the technique is really run. Others are safety-relevant: the prompt sits near dual-use territory, and you assess whether the model's refusal, hedge, or partial answer was calibrated or clumsy. A third slice is classification work — applying a written rubric to sort prompts and conversations into defined categories, consistently, across many items.

The Spanish requirement is substantive rather than cosmetic. Much of the value here is in producing scientific register in Spanish: correct IUPAC-style nomenclature, standard laboratory phrasing, terminology that a Spanish-speaking researcher would recognise rather than a literal translation of English. English is needed at business level because guidelines, rubrics, and reviewer feedback come in English.

What the screen looks for

Mercor's screening is AI-led and leans on verifiable specifics. Expect it to establish your PhD status and field, then push on subfield depth — instrumentation you've personally run, syntheses or assays you've done yourself, how you'd know a plausible-sounding answer is wrong. It also probes safety judgment: whether you can distinguish genuinely hazard-uplifting detail from information that is merely alarming-sounding but standard in the literature. Depth in one subfield reads better than shallow coverage of three. Radiological and nuclear background from a chemistry route is explicitly in scope.

Logistics

  • Remote and asynchronous; you pick up task batches rather than holding fixed shifts.
  • Start date is immediate, so realistic near-term availability matters more than long-term commitment.
  • Spain or Western Europe is stated as a preference only — applicants elsewhere are considered.
  • Pay observed at $58–62/hr; rates on Mercor vary by project and are not guaranteed.
  • No AI/ML background needed — workflow training is provided.