What the work involves

You write realistic manufacturing problems and judge how models answer them. A typical batch might ask you to draft a scenario about a CNC line drifting out of tolerance, a supplier PPAP submission with a flawed Cpk calculation, a lean kaizen event that stalled, or an OEE report where availability and performance losses have been double-counted. You then produce a reference answer, define what a correct response must contain, and score model outputs against it — flagging where the model invents tolerances, misapplies a GD&T callout, cites the wrong ISO or IATF clause, or gives advice that would fail a real audit.

Much of the value is in catching plausible-sounding errors. Models are fluent about takt time, SPC, and root cause analysis; they are much weaker at knowing when a control chart signal is a false alarm, when a 5-Why chain has jumped to a conclusion, or when a proposed fixture change would violate a safety interlock. Your job is to write items that separate surface fluency from working knowledge, and to explain your grading in enough detail that another engineer could reproduce it.

What the platform screens for

  • An AI-led interview covering your manufacturing background, then follow-up questions that push into specifics — expect to be asked about actual line rates, scrap percentages, or a corrective action you personally owned.
  • Verifiable experience. Plant, process, quality, industrial, or manufacturing engineering roles; a degree or recognized credential (PE, CQE, Six Sigma Black Belt, CMfgE) strengthens the profile.
  • Written clarity. Rubrics and error explanations are the deliverable, so writing quality is assessed directly.
  • Calibration. Sample grading tasks check whether you score consistently and can justify partial credit rather than defaulting to pass/fail.

Logistics

Fully remote and largely asynchronous. Most contributors take 10–25 hours per week against rolling deadlines, though project volume fluctuates and there is no guaranteed minimum. Onboarding usually includes a paid or unpaid calibration round before full task access. Pay is hourly or per-task depending on the project; the $50–100/hr band reflects observed rates for this category and is not a promise.