What the work actually looks like

You spend most of your time writing prompts in Indonesian that only someone with your training could write — questions that sit at the edge of a real subfield, phrased the way a graduate student, a lab technician, or a curious non-specialist in Indonesia would actually phrase them. Then you read the model's answer and score it: is the chemistry right, is the protocol plausible, did it hallucinate a reagent or a half-life, and did it handle the sensitive parts of the question the way a responsible expert would? A third slice of the work is classification — taking a written rubric and applying it consistently to prompts and conversations, including ones you didn't author.

The safety dimension is not incidental. Coverage spans chemical, biological, and radiological/nuclear subject areas, which means you will regularly encounter dual-use material: information that is legitimate in a teaching or research context and harmful when operationalised. The task is not to refuse everything technical, nor to wave everything through. It is to distinguish between the two with reasons you can write down. No prior AI/ML experience is expected — the workflow is taught — but the judgment is not teachable in onboarding, which is why the screen focuses on it.

What the platform screens for

  • Verified doctoral depth in one subfield, not survey-level breadth. Expect follow-ups that go two or three layers past your first answer — instrumentation you've personally run, why a synthesis failed, what a spectrum actually showed.
  • Genuine Indonesian technical register. Can you render specialist terminology naturally in Indonesian, including where Indonesian scientific writing borrows English terms and where it doesn't? Conversational fluency alone tends not to survive this probe.
  • Calibrated safety reasoning. Screeners present borderline requests and listen for a defensible line, not a reflex.
  • Realistic availability at roughly 7 hours per week, sustained.

Logistics

Fully remote and asynchronous, part-time at about 7 hours weekly, immediate start. Being based in Indonesia or wider Southeast Asia is a stated preference only — applicants elsewhere are explicitly welcome, provided the Indonesian fluency is native or near-native. Pay in the $19–23/hr band has been observed for this listing; rates on Mercor vary by project, assessed depth, and task type, and are not guaranteed.