AI Quality Analyst (Personalization): Thai

What you'll do

This Thai version of Turing's personalization project asks you to be both the tester and the judge. You invent realistic multi-turn prompts (usually one to five turns) from your own context, run them through Gemini, and decide whether the Thai answers were properly grounded in what the model could actually know about you.

The rubric is grounding, integration and helpfulness. Along the way you rank two responses side by side, write rationales that name the turns where things went right or wrong, pull "Debug Info" to confirm data sources, and remove the evaluation chats from your history.

Who fits

• High-level Thai reading and writing; Thai is the focus language
• Willing to use your main Google account with personal data sources turned on
• Good at noticing unnatural phrasing, forced links to your data and unsupported claims
• Structured written feedback
• Relevant degree or equivalent experience; annotation, AI quality or moderation experience strongly preferred
• Desktop or laptop and good internet

What it pays

$15/hour as the ad states, for a three-month engagement. The rate matches the Portuguese, German, Vietnamese and Turkish versions; the Dutch and Polish versions pay $20 for one month.

Worth knowing

Good:

• A three-month horizon
• Hours flex from 4 a day up to 40 a week
• Defined, rubric-based work

Less good:

• The test material is your real inbox, searches and viewing history
• The ad wants full-time availability in your local time zone plus 4 hours of PST overlap, a late-night block in Thailand
• Modest rate
• Contractor, no benefits

How hiring works

Job Interest Form, a 24-hour assessment, then a pre-onboarding conversation.

About this listing

Terms are Turing's own; the commentary is ours. See Turing.