What the work involves

You will be building and stress-testing psychiatric case material that an AI model learns from. In practice that means drafting complex presentations — mood and anxiety disorders, psychosis, developmental and geriatric cases, culture-bound presentations such as hikikomori or taijin kyofusho — and then annotating what a competent clinician would and would not conclude. A large share of the work is evaluative rather than generative: reading a model's assessment, diagnostic reasoning, or patient-facing explanation and marking precisely where it goes wrong, why, and what the correct reasoning chain looks like. Expect to write in both Japanese and English, sometimes producing parallel versions where the Japanese cannot be a literal translation because the clinical register, family involvement, or disclosure norms differ.

What the platform screens for

  • Verifiable credentials. MD/MBBS or equivalent, completed psychiatry residency, current licence. Screens ask where you trained, where you are licensed, and what you actually do clinically now.
  • Genuine bilingual clinical ability. Not conversational Japanese and not textbook English — the ability to write a formulation in Japanese and a defensible rationale in English. Screens often switch languages mid-conversation.
  • Japanese clinical grounding. Experience inside Japanese practice settings: DSM/ICD use alongside local conventions, Japanese pharmacological norms and dosing habits, involuntary admission categories, insurance and referral realities.
  • Evaluation judgment. Can you separate "wrong" from "unconventional but defensible"? Can you write a critique another reviewer could act on without talking to you?
  • Safety instinct. Sensitive content — self-harm, abuse, coercive admission — requires calibrated handling rather than blanket refusal or careless detail.

Logistics

Fully remote and contractor-based. Work arrives as batched tasks with deadlines rather than shifts, so most contributors fit it around clinical practice — evenings, weekends, or blocked half-days. Observed rates for this category sit in the $100–200/hr range, varying with credential depth, language verification, and task type; nothing here is guaranteed, and volume fluctuates with project phase. Many contributors start with a paid calibration batch before steady allocation. Patient confidentiality applies fully: cases must be constructed or fully de-identified, never lifted from your records.