What the work actually involves

You are not answering statistics questions; you are writing the exam and the answer key. A typical task starts from a realistic scenario you construct — a phase III trial with informative dropout, a propensity-score analysis of a registry cohort, a reviewer's query on a subgroup claim — and requires you to source or build the supporting material: synthetic or de-identified trial data, protocol and SAP excerpts, patient-level records, a mock CSR section. You then define the defensible methodological solution and the clinical interpretation that follows from it, and decompose your reasoning into a rubric of 35 or more discrete, checkable items covering estimand definition, missing-data handling, multiplicity, model assumptions, and whether the AI's conclusion is supportable from the evidence it had.

The explicit instruction is to avoid textbook-style problems. A task where the correct answer is "use a mixed model for repeated measures" is not useful; a task where MMRM is defensible but so is a multiple-imputation approach, and the rubric rewards a stated rationale and penalises unexamined assumptions, is.

What the screen looks for

micro1's screening is AI-led and follow-up heavy. It probes whether your stated experience holds up in specifics: which therapeutic areas, which regulatory interactions, which software applied to what kind of data. Expect questions that push past the first answer — name a method, and you will be asked what you would do when its assumptions fail. The second thing it measures is rubric judgment: whether you can turn statistical expertise into graded criteria that distinguish a partially correct analysis from a confidently wrong one, rather than a pass/fail opinion.

Logistics

  • Fully remote, asynchronous, contractor engagement.
  • Compensation is per accepted task, not hourly; the $60–100/hr band reflects observed effective rates and depends heavily on how quickly you converge on the required format.
  • A minimum weekly task submission applies.
  • Roles typically fill within about 48 hours, with first tasks expected 24–48 hours after onboarding — availability is genuinely part of the assessment.
  • No prior AI or ML experience is expected or required.