Practising Surgeon · Medical AI Evaluation · Doctor-Builder

A practising surgeon reviewing medical AI for accuracy, reasoning and safety.

I evaluate AI-generated medical answers, clinical reasoning and differential diagnosis; develop difficult cases and scoring rubrics; identify unsafe advice and missed escalation; and design clinician-in-the-loop review workflows.

I also help doctors and clinics strengthen their Google visibility and digital trust.

Explore My Medical AI Work Explore Systems for Doctors
  • More than 20 years in clinical practice
  • Practising laparoscopic surgeon, Abu Dhabi

MBBS · MS (Surgery) · FACS · FICS · FIAGES

Diploma in Laparoscopic Surgery, Strasbourg, France

Portrait of Dr Rajarshi Mitra
Two Pathways

Two ways I apply clinical judgement and systems thinking

Whether evaluating a medical answer or improving how patients find a doctor, the method is the same: examine the evidence, identify failure points, preserve human accountability and build a system that works in practice.

Medical AI Evaluation

Reviewing what AI tells patients and clinicians

  • Medical-answer evaluation
  • Clinical-reasoning assessment
  • Difficult cases and rubrics
  • Unsafe-output identification
  • Evidence adherence
  • Escalation and human oversight
Explore Medical AI Work
Digital Systems for Doctors

Helping good doctors get found and trusted

  • Google visibility
  • Google Business Profile improvement
  • Ethical review systems
  • Patient-discovery pathways
  • Digital trust
Explore Clinic Growth
Contribution Areas

Where clinical judgement changes the output

Medical-answer evaluation

Reading an AI-generated answer the way a clinician reads a referral letter — what it claims, what it leaves out, and what a patient would do next.

Clinical-reasoning assessment

Following the reasoning and the differential rather than the final answer, to see whether the conclusion was reached soundly or arrived at by chance.

Difficult cases and evaluation rubrics

Writing the awkward presentations — atypical, comorbid, ambiguous — and the rubrics that let different reviewers score them consistently.

Unsafe-output and missing-escalation detection

Identifying advice that is confidently wrong, and answers that fail to say when a patient needs urgent review.

Evidence and guideline adherence

Checking whether an answer reflects current evidence and guidance, or simply sounds authoritative enough to be believed.

Clinician-in-the-loop workflows

Designing review steps that keep a qualified human accountable for what reaches a patient, at the point where it still matters.

See how this work is structured →

Featured Medical AI Case Study

From a multi-stage content pipeline to a single AI drafting step—without removing clinical control

Designed around one principal article-generation call. Multi-step clinician-controlled release.

The original medical-content workflow separated research, claims, briefing, drafting, review and technical controls across multiple stages. It was rigorous but operationally heavy. The current Direct method is designed to consolidate the principal article-generation work into one evidence-informed call while retaining evidence artifacts, deterministic validation, clinician review, bounded correction, explicit publication authorization and recorded post-release verification.

What clinical review caught

Retained project artifacts document three clinician-directed correction cycles before publication.

  • Laparoscopic port placement
  • Biliary anatomy after gallbladder removal
  • Pain-location laterality

The redesigned method

  1. Topic boundary, evidence and governing standards assembled
  2. Principal article-generation step One principal generation call is the operating design; per-run invocation logs were not retained.
  3. Deterministic content and technical checks
  4. Clinician review and bounded correction where required
  5. Explicit publication authorization, followed by deployment and recorded live verification

What remained under clinical control

  • Every article remains subject to Dr Mitra’s clinical review.
  • Medical illustrations require his approval before publication.
  • Publication authorization remains separate from article generation.
  • Deployment and recorded live-route verification follow authorization as separate technical stages.
Doctor-Builder

Clinical experience first. Systems building followed.

More than 20 years in clinical practice came first. Everything else followed from it.

Operating lists and clinics teach a particular kind of attention: where a history is thin, where a reassuring answer is hiding an unanswered question, and where the next step actually changes what happens to a patient. That is the same attention I bring to an AI-generated medical answer.

The systems work started the same way — with practical problems. How a patient finds the right doctor. How a profile earns trust before anyone picks up the phone. How a piece of medical writing gets checked before it is published. Each one was solved because it needed solving, then turned into something repeatable.

AI sits inside that work as an assistive tool. It drafts, suggests and accelerates. It does not carry clinical responsibility, and it is not the final authority on anything that reaches a patient.

About Dr Rajarshi Mitra →

Digital Systems for Doctors

Good doctors should not lose patients because their digital presence is weak.

I built a practical Google visibility and review system for my own professional presence — working out what actually moves a profile, what patients look at before they call, and how to ask for reviews without breaking policy or trust.

Those same lessons now go to other doctors and clinics: the same method, applied to a different problem.

Explore Clinic Growth Request a Google Business Profile Health Check

Five free tools for doctors and practice managers — audit checklists, a self-audit scorecard, a review SOP and a competitor worksheet. Open the Insight Vault →

  • Google Business Profile Health Check
  • Ethical review-generation systems
  • Practical digital-trust improvements
Work With Me

Working on something where clinical judgement matters?

Tell me what you are building and where a practising clinician would be useful.

Discuss medical AI evaluation work →

Running a clinic instead? Start with a Google Business Profile Health Check →