What the work involves
You will be producing and reviewing the kind of software engineering work a frontier model still fails at: multi-file changes in real repositories, debugging traces that require reading unfamiliar code, ambiguous tickets where the correct move is to ask a question rather than write a patch. Depending on how the project is staffed at the time you join, that can mean authoring tasks with verifiable solutions, writing reference implementations and test suites, grading model attempts against a rubric, or writing the critique that explains precisely where a plausible-looking diff goes wrong. The project is expected to run several months, with internal promotion paths — task authoring to review to lead — and Mercor treats strong contribution history as a signal for future placements.
What the platform screens for
The stated bar is 5+ years at top US tech firms, and the screen is built to test whether that experience is load-bearing. A 30-minute AI interview covers your background and technical depth, then an asynchronous SWE exercise plus a project instructions quiz (roughly two hours total, paid $100 on completion). The quiz matters more than candidates expect: it checks whether you actually read specifications, because the day-to-day work is following detailed annotation guidelines without drifting into your own preferences. Expect follow-up pressure on any system you claim to have built — the interview is looking for specific decisions, tradeoffs and failure modes rather than a résumé summary.
Logistics
- Fully remote, but restricted to US and Europe-based engineers
- Around 20 hours per week; async work with periodic syncs, so hours are largely self-scheduled
- Contract engagement, expected to last several months
- $150–210/hr as observed on the listing, set by geography and experience level; not guaranteed for any individual offer