What the work actually involves

You pick up self-contained engineering tasks and complete them to a standard that can be inspected line by line. That means implementing features or fixes in unfamiliar codebases, reproducing and diagnosing bugs, refactoring for performance or maintainability, and writing reviews that explain why a change is wrong, not just that it is. Because the output feeds model training and evaluation, the reasoning trail matters as much as the diff — commit messages, test coverage, and a clear written account of the approach you rejected and why are part of the deliverable, not overhead.

Stacks vary. One task might be a Python service with a flaky integration test; the next might be a React component with a state bug or a Terraform module that fails on apply. You are not expected to know everything, but you are expected to get oriented in a strange repo quickly and know when a problem is outside your competence.

What the platform screens for

  • Verifiable delivery history. Named languages, real projects, what you personally shipped versus what your team shipped. Vague ownership claims get unpicked under follow-up.
  • Debugging method. Screeners probe how you narrow a fault — bisection, instrumentation, hypothesis testing — rather than whether you eventually found the bug.
  • Review judgment. Can you distinguish a correctness defect from a style preference, and can you say so without hedging or nitpicking?
  • Fundamentals under pressure. Data structures, complexity, and basic system design come up as follow-ups to your own examples, not as trivia.
  • Honest availability. A stated 10–20 hours over a compressed 2–3 week window, alongside a full-time job, needs to be arithmetically plausible.

Logistics

Fully remote and asynchronous — no standups, no fixed hours. The engagement is scoped to roughly 2–3 weeks with per-project assignment, so you accept or decline work as it comes. Observed pay for this listing sits in the $50–100/hr range; rates are set per project and are not guaranteed. Expect to be evaluated on your first few submissions, with continued task flow depending on quality.