What the work involves

You write troubleshooting scenarios the way they actually arrive: a vague user report, incomplete information, a half-remembered error message. Then you write the reference resolution — not just the fix, but the diagnostic reasoning that gets there, including the branches you ruled out and why. A scenario about a user locked out after a password change should make clear whether you're checking cached credentials, a stale Kerberos ticket, a conditional access policy, or an MDM compliance flag, and how you'd distinguish them.

The second half of the job is grading. You read model output against your reference and score it for technical accuracy and, critically, for user safety. Frontier models are fluent enough to produce advice that sounds like it came from a senior admin and would still destroy a user's profile, break a domain trust, or send someone into regedit with no rollback. Flagging that gap — plausible-sounding, actively harmful — is the highest-value thing you do here.

What the platform screens for

AfterQuery's screen is AI-led and follow-up heavy. Expect it to take a claim like "Active Directory experience" and push on specifics: replication behaviour, GPO precedence, what you actually saw in an environment rather than what the documentation says. Breadth across hardware, software and networking matters more than depth in one vendor stack, because the scenarios span all three. Writing quality is assessed directly from your answers — if your reasoning doesn't survive being written down, it won't survive being used as a reference answer.

Logistics

  • Fully remote, fully async — no scheduled shifts, no live calls with users
  • Minimum commitment of 10 hours per week; you choose when
  • Paid weekly via Stripe at the listed band
  • Ongoing project work rather than a fixed-length contract, with volume varying by task queue