What the work actually looks like

You receive authentic inspection data — radiographs, UT and phased-array scans, MT/PT indication photos, sometimes thermographic imagery or remote visual inspection footage — and you produce the reasoning an expert inspector would apply. That means characterising the indication, citing the governing acceptance criteria (ASME Section V and VIII, API 510/570/653, API 1104, AWS D1.1 as applicable), and writing out the disposition and the rationale behind it in language a model can learn from. Some tasks ask you to grade or critique an AI-generated interpretation and explain precisely where it went wrong — a missed lack-of-fusion call, a misapplied slag inclusion length limit, a geometry artefact read as a flaw.

The written output carries more weight here than in field work. A field report can be terse because the reader shares your context; here the rationale is the product. Expect to spell out why an indication is rounded versus elongated, why a particular code paragraph governs rather than an adjacent one, and what additional data you would need before a defensible call.

What the platform screens for

  • Certification that can be verified — active ASNT SNT-TC-1A / ACCP or ISO 9712 Level II or III, with methods named.
  • Depth under follow-up. Screens probe: the interviewer will push past a code citation to ask why that paragraph, and what changes if the material, thickness, or service class changes.
  • Calibrated judgment on ambiguity. Candidates who force certainty on genuinely marginal indications score worse than those who state the call, the confidence, and the missing information.
  • Writing quality. A sample rationale that is precise, ordered, and free of undefined shorthand matters as much as years in the field.

Logistics

Fully remote and largely asynchronous, US-based preferred. Work is project-based and delivered in batches, so weekly volume varies — contributors commonly report 10–20 hours, with peaks when a batch opens. No AI or machine learning background is expected or asked for. A reliable connection and a display good enough to read radiographic detail are practical necessities; peer discussion on ambiguous indications happens in written threads rather than live calls.