This Man Used ChatGPT and Claude to Catch Doctors' Mistakes and Save His Mother's Life

When Pratik Desai's mother was admitted to a hospital in Pune with severe abdominal pain, the diagnosis came back as acute appendicitis. Surgery was scheduled for the following morning. Something didn't feel right to Pratik — a software engineer by training with no medical background — so he did what engineers do: he built a system to check.

Using a combination of ChatGPT-4o and Claude 3.5 Sonnet, Pratik fed in his mother's complete medical history, lab results, imaging reports, and symptom timeline. The AI flagged something the attending physician had not: the symptom pattern was more consistent with a rare presentation of ovarian torsion, a condition that requires different, urgent surgical intervention.

Three Hospitals, Three Near-Misses

Pratik's mother was transferred to a specialist. The diagnosis was confirmed. The appendix was fine. What followed was emergency surgery for the correct condition — one that, left untreated, can cause permanent damage within hours.

"I'm not saying the doctors were bad. They were working with incomplete information under time pressure. The AI had the luxury of time and perfect recall. It caught what humans missed." — Pratik Desai

Over the next three weeks, as his mother recovered and was moved between facilities for follow-up care, Pratik ran the same AI-assisted review process on every new diagnosis and prescription. He caught two additional potential errors — a drug interaction that had been overlooked, and an incorrect dosage on a post-operative medication.

The Workflow He Built

Pratik has since documented his process in a detailed GitHub repository that has received over 12,000 stars in two weeks. The workflow involves structured prompting across multiple AI models, cross-referencing outputs, and a checklist for when to escalate concerns to medical professionals.

The story has struck a chord particularly in India, where doctor-patient ratios remain challenging and patients often lack the resources to seek multiple specialist opinions. AI, in this framing, becomes a form of medical equity — giving ordinary families access to a level of analytical rigor previously reserved for those who could afford it.