What the work actually involves

You are given (or select) a demanding statistical visualization and build a multi-step analytical question around it, then answer your own question with a written derivation that a reviewer can check line by line. A typical item might ask a model to read a calibration curve and quantify over-confidence in a specific probability bin, infer flow conservation through a Sankey diagram, diagnose heteroskedasticity from a residual plot, or compare partial correlations in a matrix while respecting the colour scale. The premium is on items with one defensible answer: if two competent statisticians could disagree, the item fails.

Day to day this looks less like labelling and more like exam authoring. You will write explicit calculations, name the inference you are relying on, cite axes, units, legends and annotation conventions precisely, and flag where a chart genuinely cannot support a conclusion. Reviewer feedback comes back on ambiguity, on solutions that skip an arithmetic step, and on questions that reduce to surface-level chart reading. Incorporating that feedback consistently matters more than raw volume.

What the platform screens for

micro1's screening is AI-led and conversational, followed by a domain assessment. It probes whether you can talk fluently about chart families you claim to know — what a reliability diagram shows that an ROC curve does not, when a log scale changes the correct reading, what a shaded band around a fitted line actually represents. Expect follow-up questions that push one layer past your first answer, and expect to be asked to reason aloud about a described chart without seeing it. Clear spoken and written English is assessed directly, because your explanations are the deliverable.

Logistics

  • Fully remote, contractor engagement, invoiced hourly at observed rates of $25–50/hr; bands are not guaranteed and vary by project and assessed level.
  • Largely asynchronous with written and occasional verbal check-ins; work is drawn from a queue rather than assigned in fixed shifts.
  • No prior AI or annotation experience is required — quantitative domain depth is the gate.
  • Volume fluctuates with customer demand; treat it as part-time supplementary work unless told otherwise.