Bilingual Thai Generalist – MacBook Required

What you would do

• Speak Thai voice commands to the AI system asking it to compose messages, generate images, apply formulas, or summarize documents
• Create accurate transcripts of your spoken queries, editing out stutters and filler words while keeping meaning intact
• Rate how well the AI understood and responded to each voice command on accuracy and usefulness
• Record detailed task logs in a structured tracking sheet with transcripts, ratings, and metadata
• Upload audio recordings and completed worksheets to shared storage as required by the project

Who they want

• Native or near-native Thai speaker, ideally raised or native to Thailand
• Able to understand and follow English instructions clearly
• MacBook with Apple Silicon processor (M1, M2, or M3) and current macOS Sequoia installation
• Regular daily user of productivity apps like email, messaging, documents, spreadsheets, and browsers
• Comfortable with detailed data entry and self-directed quality review of structured information

Main skills

What the interview asks about

1. 1.Voice interaction and natural language use

   The quality of AI training depends on realistic voice interactions. You need to use Thai naturally, not robotically, so the AI learns from authentic speech patterns.

   For example: “You're asking the AI to compose an email in Thai to your manager about a project delay. Do you speak formally, casually, or somewhere in between? How might you phrase this differently than writing it, and why does that distinction matter for voice AI training?”

2. 2.Transcription accuracy and editing

   Transcripts train AI systems on what was actually said versus what was meant. Your cleaned transcripts show the AI the difference between natural speech and intended text.

   For example: “You said: 'Um, hey AI, create like, I mean, format the spreadsheet with the, uh, the quarterly numbers?' How would you transcribe this for the training data, and what details matter for the AI to learn from it correctly?”

3. 3.Evaluating AI response appropriateness

   Ratings help researchers understand where AI succeeds and fails with Thai speakers. Your judgment of whether a response was accurate, helpful, or risky informs model improvements.

   For example: “You ask the AI in Thai to 'summarize this document with focus on the important parts.' The AI summarizes it but omits financial figures. You considered them important. How would you rate this response and why is that distinction worth logging?”

4. 4.Data logging consistency and precision

   Hundreds of precise task logs become the signal for training. One inconsistent or missing field can reduce data quality for the entire project.

   For example: “Over 50 tasks, you notice some entries have timestamps and some don't, some ratings are one number and some are ranges. By task 35, do you go back and fix earlier entries or establish a standard going forward? Why does that choice matter?”

5. 5.Multi-app workflow navigation

   Real users bounce between email, messaging, documents, and spreadsheets. The AI needs to handle these transitions seamlessly, and your interactions will show where it struggles.

   For example: “You ask the AI: 'Take the numbers from this email and plug them into the spreadsheet I was working on earlier.' The system retrieves the email correctly but loads the wrong spreadsheet. How would you capture this failure in your logs and rating?”

A task you may get

Use Thai voice to ask the AI to compose a meeting reschedule message, then format an attendee list. Transcribe your commands, rate responses, and log the task with supporting recordings.

How to prepare

• Practice speaking naturally in Thai about workplace tasks like scheduling and email composition, maintaining authentic voice rather than sounding scripted
• Review your Mac settings and test the app installation well before the project start so you're familiar with the technical setup
• Prepare examples of how you'd rate AI responses - what makes a response 'good' versus 'needs work' in terms of Thai language understanding and practical utility
• Develop a personal data logging template that tracks task ID, query, AI response, rating, issues encountered, and timestamps so your logs stay consistent across all 50+ tasks

The facts

Pay

$13/task

Hours

Task based, 40 hours a week

Where

Remote

Posted

9/4/2026

We wrote this page from the public Mercor listing. It may be out of date, so read the full posting before you apply.