What the work involves

You design short conversations — typically one to five turns — that force the model to reach into your actual personal context: past Gemini chats, Gmail, Google Search history, YouTube activity. Because the prompts come from your life, nobody can hand you a script. You decide what a genuine personalization opportunity looks like, then check whether the model took it, missed it, or invented something.

Evaluation runs along a few named dimensions. Grounding: is every claim the model makes about you actually supported by a retrievable data source, or is it a plausible-sounding inference? Integration: is the personal detail woven in naturally, or does the model overnarrate — announcing that it checked your email before answering a simple question? Helpfulness: side-by-side stack-ranking of two responses, with a written rationale that points to the specific turn where the difference showed up. You'll also pull Debug Info to verify which sources and chat summaries the model actually used, and delete your evaluation conversations afterwards so they don't contaminate your own chat history.

What the screen looks for

Turing's process starts with a Job Interest Form, then a profile review, then a timed assessment that must be returned within 24 hours. The assessment is where most candidates are filtered: it tests whether you can tell a flawed inference from a hallucination, whether your German rationales are structured and concrete rather than impressionistic, and whether you'll commit to using a primary personal Google account with data sources enabled. Prior annotation, content moderation, or AI quality work is strongly preferred; a BS/BA in an analytical field (linguistics, law, policy, journalism, CS) is the stated baseline, with equivalent experience accepted.

Logistics

  • Remote contractor engagement, roughly three months.
  • Minimum 4 hours/day, up to 40 hours/week, with 4 hours overlapping PST — the team runs 24-hour global coverage, so your local schedule matters less than that overlap window.
  • Desktop or laptop with reliable internet; the work is browser-based and not async in the loose sense — throughput and turnaround are tracked.
  • Observed rate on this posting is $15/hour. Turing rates vary by project and are not guaranteed across engagements.