$115–150/hr · Mercor · Part time, 40 hours a week
Board-certified physician who grades AI diagnostic reasoning against clinical expertise and real-world practice constraints.
What you would do
- Score AI-generated differential diagnoses for ordering logic and clinical appropriateness
- Validate proposed workup plans against current clinical guidelines and evidence
- Construct clinical vignettes from your practice with presentation, labs, imaging, and correct reasoning
- Identify where AI recommendations are academically sound but contraindicated by patient factors
- Document and explain gaps in AI diagnostic reasoning
Who they want
- Board-certified physician (M.D. or D.O.) holding an active, unrestricted medical license
- Current clinical practice or recent enough to maintain sound judgment
- Exceptional written communication for explaining complex medical decisions
- Able to work with ambiguous clinical scenarios and comfortable raising concerns about unclear guidance
What the interview asks about
1.Differential diagnosis ranking
Your ability to justify why one diagnosis should be prioritized over another is central to grading AI reasoning quality.
For example: “Walk me through a recent patient with chest pain, dyspnea, and elevated troponin. How did you order your differential, and what would make you move cardiac etiology down the list?”
2.Guideline-reality gaps
Knowing when standard recommendations fail for specific patients demonstrates mature clinical judgment.
For example: “Describe a case where you recommended against a guideline-standard antibiotic because of patient allergies or contraindications. How would you explain that trade-off?”
3.Anchoring bias recognition
Spotting when cognitive bias leads to premature closure separates expert judgment from AI shortcomings.
For example: “A model presents a patient with fever and focal consolidation, anchors to pneumonia, and misses signs of pulmonary embolism with infarction. What clue would you highlight?”
4.Clinical vignette writing
Creating cases with verifiable answers that reflect real attending-level judgment is essential for model training.
For example: “How would you structure a teaching case that has multiple defensible differential diagnoses, but one that a junior resident should prioritize based on presentation severity?”
5.Patient context integration
Accounting for comorbidities and access barriers is where human judgment transcends textbook medicine.
For example: “A patient with severe renal impairment and sepsis needs a certain antibiotic, but standard dosing causes toxicity. How do you evaluate an AI's proposed alternative?”
A task you may get
Review a clinical case with AI-generated differential diagnosis and workup plan; grade reasoning, flag guideline deviations, and explain real-world constraints that affect the recommendation.
How to prepare
- Review recent patients where your final diagnosis differed from initial clinical suspicion
- Document 2-3 cases where comorbidities or access constraints changed your treatment approach
- Prepare examples of cognitive biases (anchoring, availability, confirmation) you've observed in practice
- Reflect on how you would explain diagnostic logic to a medical student
The facts
- Pay
- $115–150/hr
- Hours
- Part time, 40 hours a week
- Where
- Remote
- Field
- Medicine
- Role type
- Talent network
- Posted
- 9/15/2026
We wrote this page from the public Mercor listing. It may be out of date, so read the full posting before you apply.