What the work actually involves

This is technical peer review of AI training and evaluation tasks, not design production. A task might present a data-center shell on a sloped site with a geotechnical report, a partial structural package, an equipment loading schedule for rooftop CRAC units, and a set of AHJ comments — and ask a model to resolve a foundation or code question over many steps. Your job is to work the problem as an engineer would, then judge whether the task is sound: are the governing load combinations right, does the drawing set contradict the specification, is document precedence stated, is the expected answer actually the only defensible one? Where the task is broken, you repair it — tightening the prompt, correcting artifacts, rewriting acceptance criteria and rubrics, and adding programmatic checks — then validate the corrected version end to end.

What the platform screens for

Mercor's screening is AI-led and interview-style, with follow-up questions that go deeper when your first answer is thin. Expect to be pushed on specifics: which code editions and standards you work from, how you handle wind and seismic criteria on a large low-rise box, what geotechnical inputs drive your slab and foundation decisions, how rooftop equipment loads and vibration get coordinated with the structural frame, and how you resolve conflicts between drawings, specs and addenda. Naming projects, jurisdictions, criteria and the reasoning behind a decision reads far better than describing your title. Reviewers who can also articulate why an evaluation item is unfair or under-specified — separate from whether they know the answer — stand out, because that is the actual output of the role.

Logistics

  • Fully remote and largely asynchronous; work is picked up from a queue rather than scheduled in meetings
  • Roughly 5–10 hours per week initially, with immediate start and potential to scale
  • Observed compensation of $100–$140/hr, set by experience and role fit — a band, not a guarantee
  • Written deliverables dominate: technical findings, redlines to task text, rubric revisions

The strongest fits tend to be Engineers of Record, senior structural or civil engineers, code consultants, project architects and BIM/VDC managers who have carried mission-critical work through permitting and construction and can explain a technical position in writing to someone who was not in the room.