What the work involves

You write and grade tasks that put a model's clinical reasoning under pressure. In practice that means constructing prompts from trial-adjacent material — protocol excerpts, investigator brochures, CSR sections, safety narratives, DSMB/IDMC packets — and then judging whether a model's output holds up clinically. Typical questions: does the stated effect size actually support the efficacy conclusion for this endpoint? Is the AE grade consistent with CTCAE and with the narrative's own description? Is attribution to study drug defensible, or is the model asserting causality the data won't carry? Does the dose modification follow the protocol, and would a real investigator have done it?

A second, equally important half of the job is written rationale. Every judgment needs a structured explanation that a non-clinical reviewer — an AI trainer, not a colleague — can follow and reuse. Reviewers who can say "this is wrong" but not "this is wrong because Lugano requires X and the report applied Y" don't last long on this kind of program.

What the screening measures

micro1's process is AI-led and conversational, with follow-ups that go deeper when you make a specific claim. Expect it to probe:

  • Verifiable credentials — specialist registration or board certification in medical oncology and/or hematology, plus 5+ years post-training.
  • Trial-side experience — investigator, sub-investigator, medical monitor, safety physician, or clinical development physician; committee work (IDMC, DSMB, adjudication, IRC) is a strong signal.
  • Framework fluency under follow-up — CTCAE grading, RECIST/Lugano/IMWG/IWCLL/ELN, and what ORR, DoR, PFS, OS, and MRD do and don't tell you.
  • Evaluation judgment — whether you can separate a genuinely implausible claim from legitimate practice variation, and whether you flag with reasons rather than instinct.

No prior AI or annotation experience is required or expected.

Logistics

Fully remote, contractor engagement, asynchronous. Most clinicians on comparable programs work 5–15 hours per week on their own schedule, with occasional calibration sessions or written guideline updates. Volume is project-driven and can fluctuate between cohorts, so treat this as a supplement to clinical practice rather than a replacement. Pay of $160–200/hr reflects observed rates for this listing and is set per project, not guaranteed.