What the work actually involves

The unit of work is a research problem or a graded model response, not a shift. On authoring tasks, you take a system you have genuinely worked in — an organism, an assay, a dataset — and write it up with the controls and the reference answer a model's attempt will be scored against. On grading tasks, you read a model-written experimental design or figure interpretation and decide whether the controls rule out what they claim to, whether the statistics fit the design, and whether the stated conclusion actually follows from the data shown. A third, harder category is bench-reality review: the model produces an answer that would pass a qualifying exam but collapses given the doubling time of the organism, the dynamic range of the assay, or what a particular antibody can actually resolve. Most of your output is written prose explaining why, so the reasoning is the deliverable.

What the screen is looking for

A graduate degree in biology or a related life science with real research experience behind it. The listing is explicit that where you trained matters less than whether you designed and ran the experiments yourself, recently enough to defend the calls you made — so the AI interview will push on specifics: which organism, which construct, what the negative control was and why. Clear written reasoning matters as much as the knowledge, since a grade with no defensible rationale is useless as training signal. They also screen for tolerance of ambiguity: tasks arrive underspecified and the expected behaviour is to flag the gap, not to quietly guess.

Logistics and what to expect

  • Fully remote and asynchronous; you confirm your work location at application.
  • Application is resume plus a roughly 20-minute AI interview, after which your background is verified and you enter the biology pool.
  • This is a standing listing, not an opening — it issues no accept or reject. Matching invitations arrive when a client project needs biology, which the listing says can be within a week or several months.
  • Rate, hours, and client are named on the specific project listing, where the actual hiring decision happens.
  • Observed past rates have run $70–105/hr, set per project by scope and depth; nothing here is a guaranteed band or a guaranteed volume of work.