What the work involves

You'll be authoring and critiquing psychiatric content that an AI system learns from. In practice that means constructing complex case vignettes — presentation, history, differential, management plan — that reflect how psychiatry is actually practised with Marathi-speaking patients, then writing them up in both languages. A second stream of work is evaluation: reading model-generated clinical reasoning or patient-facing explanations and judging whether the diagnosis, the risk assessment, and the cultural register hold up. You will often be asked to explain why an output is wrong, not just flag it, because the rationale is what trains the model.

Cultural nuance is the substance of the job, not a garnish. Somatic presentations of depression, family-mediated consent and treatment decisions, idioms of distress that don't map cleanly to DSM or ICD categories, stigma-driven help-seeking delays, the role of faith healers and traditional practitioners — these are the things a general English-language dataset gets wrong and the reason the project needs you. Equally important is terminology: much Marathi psychiatric discourse borrows English terms, and part of the work is deciding when a Marathi rendering is genuinely used by clinicians and patients versus when it would read as invented.

What the screening looks for

  • A verifiable medical degree (MBBS/MD or equivalent), completed psychiatry residency, and a current licence.
  • Real clinical hours with Marathi-speaking patients, not just conversational Marathi.
  • Writing that stands up in both languages — screens frequently ask for a short bilingual sample or a live explanation of a clinical concept in Marathi.
  • Evaluation judgment: whether you can separate a confidently-worded but clinically unsafe answer from a terse but correct one, and articulate the difference.
  • Ethical instinct around de-identification. Case material must be synthetic or fully de-identified; a candidate who offers real patient records is a problem, not an asset.

Logistics

Fully remote, contractor engagement, largely asynchronous with occasional calls with project teams across time zones. Most contributors work part-time alongside clinical practice — commonly 8–20 hours per week — and volume moves with project phase rather than being guaranteed. No AI background is expected; the onboarding covers the annotation interface and rubric. Pay in the $100–200/hr band has been observed on comparable micro1 medical-specialist projects and varies with seniority, language pair, and task type.