What the work involves

You write the problems, not the code. A typical task starts with a realistic civil engineering situation — a highway interchange geometry revision under AASHTO criteria, a culvert sizing check against a revised HEC-RAS floodplain model, a settlement analysis for a shallow foundation on compressible soils, a schedule and cost recovery plan after a differing site condition claim — and you build it out with the inputs an engineer would actually have: soil boring logs, survey data, design criteria memos, agency comment letters, submittal logs. You then produce the reference output: the calculation, the plan narrative, the RFI response, the permitting strategy. Finally you write a rubric that says precisely what separates a defensible answer from one that recites code sections without applying them.

Rubric authorship is where most of the difficulty sits. Models are fluent in ASCE 7 load combinations and NEPA process vocabulary; they are much weaker at knowing when a design assumption is unconservative, when a regulatory sequence is out of order, or when a constructability issue makes a technically valid design unbuildable. Good rubrics make those distinctions explicit and checkable by a reviewer who wasn't in your head.

What the screen looks for

  • Ownership, not exposure. Mercor's screen probes for projects you personally stamped, designed, or delivered — scale, delivery method, your specific scope. Vague program-level involvement tends not to survive follow-up.
  • Depth under pressure. Expect the AI interviewer to pick one discipline you claim — structural, geotechnical, water resources, transportation, or construction management — and push two or three questions deeper on assumptions, failure modes, and standards.
  • Regulatory realism. Understanding how environmental review, agency comment cycles, and permitting actually sequence in practice is weighted heavily, because it's a common model failure mode.
  • Evaluation judgment. Can you articulate why one answer is better than another in terms someone else can apply consistently?

Logistics

Remote and fully asynchronous. Contributors typically commit 10–20 hours per week, though some engagements run heavier during ramp-up. Work is submitted in batches with reviewer feedback loops, so expect revision cycles on your first several tasks while you calibrate to the rubric standard. PE licensure is strongly preferred and generally strengthens both acceptance odds and rate placement within the observed $70–80/hr band. Pay is as observed on the platform and not guaranteed.