Clinician AI Task Author / Reviewer ($50-$150 per task), Remote
ABOUT THE WORK We build realistic simulations of clinical work that are used to train and evaluate AI systems. Each simulation is a task packet: a request the way an attending, a referring colleague or a nurse would actually make it, the chart a clinician would really have (presenting complaint and history, medication and allergy lists, vitals, labs and imaging reports, prior notes, discharge summaries, referral letters), an answer key, and a grading rubric that says what a correct answer must contain, what would be unsafe, and what is merely style. We need practising clinicians for two kinds of work. Reviewing: read a packet we wrote, say where the chart or the answer key is wrong, and mark up a short form. Authoring: take a kind of case you have managed many times, disguise it, build the chart, write the answer key, and design the rubric a senior colleague would grade it against. Most people start with reviewing and move to authoring once their first packets are good. Specialty is open: internal medicine, family medicine, emergency medicine, hospital medicine, pediatrics, psychiatry, OB/GYN, cardiology and others. We care more about depth in one specialty than breadth. All patient data is synthetic; you never share real patient information. We are especially interested in clinicians who have already worked on AI training or evaluation projects. WHO WE NEED Must have: - 3+ years of hands-on clinical practice as a physician (MD/DO), PA or NP, currently practising or within the last 2 years - Have personally worked up patients and written the assessment and plan yourself, not only reviewed other people's notes - Can say precisely why a clinical answer is wrong or unsafe, in writing, in a way a resident would understand and an attending would accept - Comfortable writing to a template and getting line-by-line feedback from researchers who are not clinicians Nice to have: - Prior work writing or grading medical tasks for AI training or evaluation - Have turned your work into something teachable: a teaching case, an M&M presentation, board-style questions, an order set, a clinical pathway, a chart-review or QA form - Have supervised or precepted residents, students, PAs or NPs - Epic, Cerner or other EHR experience; peer review or utilization review experience WHAT YOU WILL DO 1. Review task packets, about 30–60 minutes each: rate the realism of the chart and the request, flag anything a clinician would not see or would not be asked, and confirm or dispute the answer key with a one-line reason. 2. Write task packets from your own experience to our template: the request as it would really arrive, the chart a clinician would have on day one (including realistic noise: a copied-forward note, a med list that doesn't match the pharmacy fill history, a result that is still pending), the answer key, and the reasoning behind it. Typically 2 hours each. All patients and records are fictional. 3. Design the rubric for each task you write: what a correct answer must contain, what would be unsafe or get it sent back, and what is a judgment call. Scored, not prose. 4. Grade blind: review answers written by other authors or by our systems against the rubric, and say where the rubric itself is wrong. SCREENING A short recorded interview with our AI interviewer (about 20 minutes, on your own time, nothing to schedule). Please answer the three application questions. PAY, HOURS AND LOCATION $50-$150/ task based on the task. US-licensed or US-practising preferred; other jurisdictions considered if you have done AI training-data work before.