What the work actually involves

You receive model outputs — a draft Phase I ESA data gap discussion, a groundwater fate-and-transport narrative, an air permit applicability analysis, a CERCLA remedy screening — and judge them against what a practicing professional would accept. That means checking whether the regulatory citation is the right one and current, whether the conceptual site model is coherent, whether the unit conversions and screening levels survive arithmetic, and whether the recommendation would actually clear a state agency reviewer. Alongside grading, you write demonstration tasks: original problems with an expert-authored solution and a rationale explaining why the correct path is correct and why the plausible wrong paths fail.

Ranking work asks you to order several model responses and say precisely what separates them. The rationale is the deliverable — "this one is better" without a defensible technical reason is not usable training signal. Expect your written justification to be read more carefully than your score.

What the platform screens for

  • Verifiable practice history. Which statutes and programs you worked under, in what role, and what you personally produced — not what your firm did.
  • Depth under follow-up. Screens push a second and third layer: not "do you know RCRA" but the distinction between a listed and characteristic waste in a specific scenario, or when a NEPA categorical exclusion no longer applies.
  • Calibration. Whether you can distinguish a fatal technical error from a stylistic preference, and whether you flag confident-sounding fabricated citations rather than reading past them.
  • Writing. Rationales must be legible to a reviewer who is not an environmental professional but is technically literate.

Logistics

Fully remote and fully async — no standing calls, no fixed hours. Contributors typically commit 10–20 hours a week and work in blocks against a task queue with turnaround deadlines rather than clock-in windows. Pay is hourly; the $70–130 band reflects rates observed on this platform and varies with credentialing, specialty scarcity, and task complexity. Nothing here is guaranteed, and volume can fluctuate between project cycles.