What the work actually is

You pull a task from a pre-validated pool: a Python repository pinned to a specific git commit, an engineering question, positive and negative rubrics, foils, and a golden solution. Your job is not to answer the question — it's to make it harder and cleaner. That means reading the subsystem until you genuinely understand how it behaves at runtime, tracing the PR and issue history that explains why it behaves that way, then rewriting the question so a frontier coding agent fails it while a competent senior engineer with repo access can still resolve it fairly. Most tasks live in architecture and system design (roughly 41% of the mix) or code onboarding (31%), so the questions tend to be about how components interact, what happens on a particular call path, or why a design constraint exists — not about a single function's return value.

The validation loop in Mercor Studio is Check, Golden, Run, Analyze, and you iterate until every stage passes. You then write the task README and submit. A second engineer and a super reviewer both look at it before acceptance, and returned tasks come back with feedback you're expected to address. Pay is $130 per approved task as observed on the platform; approval is the gate, so rework is part of the economics.

What the screen is looking for

  • Runtime reasoning, not source reading. The distinguishing signal is whether you can explain what a Python system does when it runs — import order, descriptor and metaclass behaviour, event loop scheduling, packaging and entry points, the C-extension boundary — rather than paraphrasing the file.
  • Evidence of navigating unfamiliar large codebases. Concrete stories about landing in a repo you didn't write and finding the answer.
  • Adversarial instinct. Can you predict how a strong coding agent would shortcut a question, and close that path without making the question unanswerable?
  • Written precision. The questions and rubrics are the deliverable. Vague prose fails review.

Logistics

Fully remote and asynchronous, with expert team leads spread across US, India and Nigeria time zones. Work happens inside Mercor Studio; time is tracked with Insightful and access is provisioned through Okta. Onboarding requires identity verification, a background check, signed Terms of Work and CIIAA, a tax form, and a short calibration quiz before task access opens. There's no fixed schedule, but tasks are substantial — expect several hours of genuine investigation per task, not a quick turnaround.