What the work actually involves

You'll be handed artifacts you already read for a living — machining drawings with GD&T callouts, CNC setup sheets and program listings, construction plan sets, drilling or mine plans, fire and life-safety submittals — alongside an AI model's interpretation of them. Your job is to judge that interpretation against how the job is really done: is the datum reference frame read correctly, would that toolpath actually clear the fixture, is the plan sheet missing a dimension a crew would call about on day one, does the egress width claim survive a look at the applicable code section. Deliverables are usually written: a verdict, a ranked account of the errors, and a corrected explanation a competent tradesperson would accept. Some tasks add recorded verbal walkthroughs of a drawing or a field scenario.

What the screen is looking for

  • Depth that survives follow-up. The screener will push on specifics — which tolerance zone, which code edition, which setup you'd actually run. Broad statements about "years in the shop" collapse quickly.
  • Citation instinct. For fire and building work, pointing to IBC/NFPA sections rather than general safety language. For machining, datums, feature control frames and material condition modifiers rather than "the tolerance looks tight."
  • Writing. The output is prose. Experts who know the answer but can't lay it out in a structured paragraph don't pass.
  • Calibration, not perfectionism. Flagging every cosmetic issue on a plan sheet at the same severity as a missing dimension reads as poor judgment.

Logistics

Contractor engagement, fully remote, US-based preferred because much of the code and standards work is US-specific. Work is asynchronous and drawn from a task queue; most contributors take on a set number of hours per week rather than a fixed schedule, and volume fluctuates with customer demand. Pay is observed in the $55–100/hr range depending on specialty and depth — machining/GD&T and code-review specialisms tend to sit higher — and is not guaranteed for any given project. Expect a short paid or unpaid calibration exercise before live tasks begin.