What the work actually is
You author cloud scenarios that a frontier model has to reason through: a VPC peering design that has to satisfy a compliance boundary, a Terraform state conflict during a multi-account rollout, a 3am pager on an autoscaling group that won't scale, a cost review where the obvious fix is a reserved instance and the correct fix isn't. Each scenario ships with a reference answer that explains the architectural reasoning, not just the resulting config. Then you grade model output — and the grading is where the judgment lives, because models produce cloud configurations that parse cleanly, apply successfully, and still leave a bucket public or triple the bill at scale.
What the platform screens for
AfterQuery's screen is AI-led and follow-up heavy. Expect it to push past the résumé line into a specific incident you handled, a specific IaC module you wrote, and a specific cost decision you made and could defend. Naming services is cheap; the screen is looking for whether you can say why a NAT gateway per AZ versus a shared one mattered in your particular topology, or what you'd check before believing a model's claim that a policy is least-privilege. Written clarity counts as a technical skill here — your reference answers become training signal, and vague reasoning is worse than no reasoning.
Logistics
- 100% remote, fully async — no standups, no fixed hours, no timezone requirement
- Minimum 10 hours per week; you set the schedule and work in batches
- Paid weekly via Stripe; the $85–165/hr band is as observed and varies with task complexity and calibration performance
- Ongoing project work rather than a fixed-term contract, with volume that fluctuates by cloud domain