The work

You are given (or you source and describe) a financial or operational visualization and you build a task around it. A task is three things: a question that cannot be answered by glancing at the chart, an answer that is unambiguously correct, and a written derivation showing every step — which gridline you read, which two points you interpolated between, what units you converted, how you got from the raw reading to the final figure. The point is to produce items where a model either gets it right or is demonstrably wrong, with no room to argue.

Good items tend to be multi-step: read two values off a candlestick series and compute a percentage drawdown; compare the slope of a yield curve at two tenors and infer the implied forward; determine whether a control chart point violates a Western Electric rule and explain which; compute cumulative contribution from a Pareto chart and identify where the 80% line falls. Weak items ask what the highest bar is. Expect feedback cycles where a reviewer pushes back on ambiguity — an unstated timeframe, a log-scale misread, a question with two defensible answers — and you revise.

What the screen looks for

micro1's process is AI-led and heavy on domain follow-up. Screeners probe whether you actually read charts for a living or only recognize them: units and scale conventions, what a secondary axis does to a comparison, why a drawdown from a log-scaled equity curve is not read the same way as from a linear one, what a widening control chart band means about the process versus the measurement. They also check written English under pressure, because your explanation text ships as training data. Prior AI or annotation experience is explicitly not required.

Logistics

  • Fully remote contractor engagement, invoiced hourly; observed band for this listing is $25–50/hr, set by assessed depth and not guaranteed.
  • Largely asynchronous with occasional calls for calibration and feedback; project teams work in English.
  • Volume is project-driven — contributors commonly report part-time commitments of 10–20 hours a week, with throughput expectations stated per batch.
  • You need reliable access to a desktop environment where you can read charts precisely and do calculations you can document.