What the work involves

You'll spend most of your time in two modes. In the first, you author reference artifacts from a prompt: a production capacity model in Excel with takt time, OEE assumptions and a bottleneck rollup; a multi-level bill of materials with revision control and make/buy flags; a root cause report that actually closes the loop from problem statement to containment, corrective action and verification. In the second, you're handed a model-generated version of the same artifact and asked to judge it — flag the errors, rank two candidate outputs against each other, and write the rationale that explains why one is better in terms a non-manufacturing annotator could follow.

The subtle failures are the point. A model will produce a capacity model whose formulas reference the wrong cell, a PM schedule with intervals that ignore duty cycle, an inspection checklist that lists attributes with no acceptance criteria, or a supplier qualification package that cites a PPAP level it then doesn't satisfy. The lab wants those caught and named precisely, not summarized as "needs improvement."

What the platform screens for

  • Ownership, not exposure. Ethos probes for workflows you personally ran — the capacity model you maintained, the corrective action you signed, the supplier you qualified — and follows up on specifics. Advisory-only or purely academic familiarity tends not to survive the follow-up.
  • Regulated-environment fluency. Working knowledge of ISO 9001 / AS9100 / IATF 16949-style expectations, 8D or DMAIC discipline, and what a supplier qualification package contains is assumed rather than taught.
  • Artifact craft. Genuine spreadsheet ability (structured models, not flat tables), document structure, and deck logic. Written rationale quality is evaluated directly.
  • Calibrated judgment. Whether you can separate a stylistic preference from a defect that would fail an audit.

Logistics

Fully remote and fully async — tasks are queued and you pull them on your own schedule. Flexible at 5–20 hours per week, with more available if you want volume. The interview includes an AI voice screen; expect roughly 20–30 minutes of spoken questions with live follow-ups on whatever you claim. Pay is observed at $90/hour and is not guaranteed; rates on these programs vary by project and task type.