What the work actually involves

You will spend most of your time doing two things: authoring reference artifacts that show the model what competent operations work looks like, and grading model output against the standard you'd apply to a direct report's deliverable. A typical task might be a staffing capacity model for a fulfillment site with seasonal demand, an SOP playbook for a returns process, a swimlane map of an order-to-cash handoff, or a business case for automating a manual reconciliation step. You write the prompt-facing version, then explain in writing why a particular formula, sequencing, or slide structure is correct — the rationale matters as much as the artifact, because that's what the model learns from.

Evaluation tasks run the other direction. You receive a model-generated workbook or deck and score it on dimensions like arithmetic soundness, assumption transparency, whether the process map actually reflects a workable handoff, and whether an SOP could be executed by someone who has never done the job. Most failures are not obvious errors — they are plausible-looking forecasts with a broken seasonality assumption, or a checklist that reads well and omits the one gate that matters.

What the platform screens for

  • Specific operational history. Ethos probes for named systems, real cycle-time or cost deltas, and the actual mechanics of how you built a capacity model — not job titles.
  • Spreadsheet depth under follow-up. Expect to be asked how you'd structure a driver-based forecast, handle assumption sheets, or stress-test a model someone else built.
  • Written judgment. Your rationale text is the product. Vague critique scores poorly even when the verdict is right.
  • Honest availability. Weekly hours you can actually sustain matter more than a large number.

Logistics

Fully remote and fully asynchronous — there are no standing meetings and no fixed hours. Throughput ranges from 5 to 20 hours per week, with more available to reliable contributors during active project windows. Screening includes an AI-led voice interview covering background, domain depth, and availability. Pay of $70/hour reflects rates observed on this platform for this role and is not a guarantee.