What the work involves

The role splits into two streams that alternate depending on project load. In the authoring stream, you write upper-undergraduate and graduate-level biology and health problems along with clear, fully worked solutions — and crucially, you convert concepts that normally lean on diagrams, gels, pedigrees, pathway maps or imaging into self-contained text-only tasks that remain solvable without a picture. In the review stream, you audit tasks written by other contributors or generated by models: checking scientific correctness, terminology precision, whether the stated answer is actually the only defensible one, and whether the item hits the intended difficulty band rather than merely sounding advanced.

A third thread runs through both: diagnosing how frontier models fail on your material. That means noting whether a model got an item wrong through a retrieval gap, a multi-step reasoning break, a plausible-sounding confabulation, or an ambiguity you left in the prompt — and then refining the item so it isolates the failure instead of blurring it.

What the screen looks for

  • A Master's or PhD in biology, health sciences, medicine or a closely related field, and the ability to talk about your subfield at depth under follow-up rather than at survey level.
  • Concrete prior experience in teaching, grading, university exam writing, or research-based analytical work — the screen will ask what you wrote or marked, at what level, and how you set difficulty.
  • Written English precise enough to survive peer review: unambiguous stems, correct nomenclature, no hedging in the worked solution.
  • Evaluation judgment: can you tell a hard question from a badly worded one, and can you defend a rejection with a specific scientific reason?

Familiarity with AI training data conventions is preferred, and prior Turing evaluation work is a plus but not required.

Logistics

Remote contractor engagement, roughly four weeks, part-time accepted. Expect a commitment of at least four hours per day, up to 40 hours per week, with four hours overlapping PST, plus weekend on-call availability — say yes to that only if you mean it, because it is an explicit requirement rather than a nicety. You supply your own desktop or laptop and a stable high-speed connection. Pay is undisclosed on this listing; rates on comparable Turing science evaluation work are typically set per hour at offer stage and vary with credential and stream.