What the work actually is

You write coding problems for AI models to attempt, and then you write the grader. A task might be a concrete bug in a service layer, a feature that has to be threaded through an existing API, or a vulnerability that needs remediation without breaking callers. The harder half of the job is the verifier: a deterministic test harness that accepts every reasonable valid solution — different variable names, different ordering, a different-but-equivalent algorithm — and rejects everything that is subtly wrong. Verifiers that only pass your own reference implementation get rejected, and so do verifiers a model can satisfy by hardcoding outputs or stubbing the function under test.

Security background is where this listing differs from a generic backend authoring role. Tasks grounded in OWASP/CWE categories — injection paths, broken access control, unsafe deserialization, SSRF, secret handling, dependency risk — are valuable precisely because they are hard to verify with naive assertions. Knowing how to prove that an injection fix actually holds, rather than that one payload stopped working, is the skill being bought.

What the screen looks for

micro1's pipeline is: screening questions, an AI-led interview of about 30 minutes, a tentative technical assessment, then hiring manager review. The interview probes language-level and systems specifics under follow-up — expect to be asked why, then why again, on debugging and testing choices you describe. They explicitly ask that you complete the interview and assessment without outside assistance, including AI tools; the project pays for your judgment, so demonstrating it unaided is the point. Vague answers about "best practices" collapse quickly under a second question.

Logistics

  • Contractor, remote, roughly 15 hrs/week, hours and days entirely your choice including weekends
  • Output-based pay: per accepted task, with a weekly minimum submission requirement
  • Roles typically fill within 48 hours; first tasks expected 24–48 hours after onboarding
  • No prior AI or ML experience required