What the work involves

You are handed (or source) a financial or operational visualization and asked to build a task around it that a strong model should struggle with but a competent analyst can verify. That means reading the chart carefully — axes, units, log vs. linear scales, legends, annotations, timeframes — then designing a question that requires two or more analytical steps: computing a peak-to-trough drawdown, comparing implied yields across tenors, inferring throughput from a capacity chart, identifying which category crosses the 80% cumulative line on a Pareto plot, or judging whether a control chart signals a special-cause shift.

Each submission pairs the question with a single unambiguous answer and a written rationale that shows the arithmetic and the reasoning path. Ambiguity is the main failure mode: questions that depend on eyeballing, on subjective judgment, or on data not actually present in the image get rejected. Precise notation, stated units, and explicit assumptions are the whole craft here.

What the platform screens for

  • Chart literacy under follow-up. Expect probes on candlestick anatomy, curve inversion, drawdown mechanics, Western Electric rules, and how log scaling changes what a slope means.
  • Verifiability instinct. Can you tell a well-posed quantitative question from one with two defensible answers?
  • Written English quality. Explanations are the deliverable; the screen reads them as such.
  • Credible domain background in finance, economics, business analytics, industrial engineering, or operations — two years or more of hands-on chart interpretation.

Logistics

Fully remote contractor work, asynchronous, with output measured in accepted tasks rather than logged hours. Most contributors set their own schedule and stay in the 10–20 hour range per week, with occasional synchronous check-ins or calibration rounds when reviewers push back on task quality. Prior AI experience is not required; project volume varies with the customer's needs, so treat the engagement as flexible rather than fixed.