The work

You write coding problems that a strong language model is likely to get wrong, then judge whether it actually got them right. A task might be a debugging scenario where the bug is a race condition rather than a syntax error, an API design prompt with conflicting constraints, or an optimization question where the naive solution passes tests but not the complexity bar. Each item usually needs a clear prompt, a reference solution, and an explanation of why plausible-looking wrong answers are wrong — that last part is often where the real value sits.

The grading half is equally substantial. You read model output that compiles, reads confidently, and is subtly incorrect: an off-by-one at a boundary, a mutation of shared state, a test suite that asserts nothing meaningful. Your feedback has to name the defect precisely and point to the principle behind it, not just mark the answer wrong. Reviewers on the platform frequently push back on vague rationales, so expect revision cycles early on.

What the screen looks for

AfterQuery's screening is AI-led and follow-up heavy. It confirms your enrollment or recent graduation in CS, software engineering, or a related technical field, then probes actual depth: it will take a claim about a language, a data structure, or a project you built and ask second and third questions about it. Expect to be asked to reason about complexity, to explain a debugging process end to end, and to judge two model answers where one is more convincing but less correct. Verifiable artifacts — a GitHub repo, an internship, a specific course project — carry far more weight than adjectives.

Logistics

  • Fully remote and asynchronous; no fixed shifts, and the role is explicitly built around class schedules.
  • Contributors typically commit a defined weekly block (often 10–20 hours) and are paid hourly on delivered, accepted work.
  • Observed rates run $40–60/hour depending on project and demonstrated depth; not guaranteed.
  • Work volume comes in project waves, so availability windows and responsiveness matter more than raw hours.