What the work involves

This is a trainer and quality role, not a pure annotation seat. You sit inside a functional pod producing training and evaluation data for Computer-Using Agent models — data that captures how an agent navigates real desktop and browser applications, which clicks and keystrokes count as correct, and where a trajectory goes wrong. Day to day you run onboarding sessions for new annotators on project guidelines, write and version SOPs and rubrics, sample and review completed annotations, score annotator performance against quality metrics and SLAs, and give feedback specific enough that the next batch improves. You also support pilot runs and calibration exercises, surface process gaps, and escalate rubric ambiguities to QA leads and project managers rather than silently interpreting them yourself.

What the platform screens for

Turing's screen is looking for someone who has actually sat inside an annotation operation, not someone who has read about one. Expect probing on concrete workflow experience: what guidelines you authored, how you measured inter-annotator agreement, how you handled a labeler whose throughput was fine but whose accuracy was drifting. CUA-specific exposure — annotating or reviewing agent trajectories across GUI applications — is a preference, not a hard gate, but candidates who can describe screen-recording, action-log, or step-level labeling work will read stronger. Written communication is assessed directly, since the job's main artifact is documentation other people must follow without you in the room.

Logistics

  • Fully remote, contractor engagement with no medical or paid leave; task-based pay, band undisclosed by the platform.
  • 40 hours per week for a stated duration of two weeks, with roughly four hours per day overlapping US Pacific time.
  • You supply your own machine: 16GB RAM minimum, Windows, Linux, or macOS, stable high-speed connection, and the ability to run project tooling securely.
  • Application flow is a Job Interest Form for shortlisted candidates, then onboarding steps for those selected. Short engagements like this are often extended when a pod performs, but nothing in the listing promises that.