For doctors

Medical AI is being trained by physicians. Not exclusively, but structurally. Every major frontier lab now runs programs where practicing doctors review, correct, and rank model outputs on clinical vignettes, diagnostic reasoning, drug interactions, differential workups, and edge cases the training data does not cover well.

The reason is simple. Medical reasoning is dense, high-stakes, and full of tacit knowledge that does not appear in textbooks. A model that reads every medical journal ever published still needs a physician to tell it when its answer is confidently wrong, when it is right for the wrong reason, and when the safer clinical move is the one that looks worse on paper.

What the work looks like

Most opportunities fall into one of a few shapes:

• Response evaluation. You are shown a clinical prompt and two or more model responses. You rank them, explain your reasoning, and flag anything unsafe.
• Question authoring. You write the kinds of clinical vignettes a resident would see on a licensing exam, at the level of specificity a model needs to learn from.
• Rubric review. You review the criteria being used to evaluate model outputs and flag anything that would misjudge good clinical practice.
• Adversarial testing. You try to elicit unsafe or wrong answers from the model, and document what worked. This is where clinical experience compounds fastest.

The work is asynchronous. You choose when you sit down. Sessions typically run in blocks of two to four hours. Reporting is structured but not oppressive.

What labs actually look for

Medical licensure. Postgraduate training in a specialty. A clear track record of clinical practice. Fluency in written English at the level a peer-reviewed journal expects.

Region does not matter. Passport does not matter. Institutional prestige does not matter as much as recent, active clinical work. A general practitioner running a busy clinic in Lahore is often a stronger candidate than a former academic who has not seen patients in five years.

Rates are set by the labs. They compare well to Western clinical compensation on an hourly basis, and are paid in USD.

How to enter

If you are a physician, sign up, upload a recent CV, and let the platform surface the roles currently open in medical annotation, medical evaluation, and clinical adversarial testing. If the fit is honest, apply through the referral link. If it is not, do not.

We are especially interested in physicians in Pakistan, India, Egypt, Nigeria, the Philippines, and other markets where this world has not been introduced yet. It should have been introduced years ago. It was not. We are correcting that.