An AI care team for a cardiology practice
A patient assistant built on the cardiologist's own material, agents that handle social comments in English and Marathi, and a dashboard where the care team reviews, edits and escalates. Medical questions always go to the doctor.
The situation
A cardiologist-led preventive care platform in Pune publishes daily on Facebook, YouTube and its website. Questions arrive faster than a small team can answer them — and some are medical questions only a doctor should answer.
Constraints we designed around
- Thousands of comments across Facebook and YouTube, in English and Marathi.
- Medical questions must never receive an automated answer.
- Every reply goes out in the practice's name, so tone and accuracy matter.
What we built
Patient assistant
A website assistant that answers from the doctor's own videos and documents. Three free questions, then phone OTP verification, then a path to book a consultation.
Comment agents
Scheduled agents read new comments, classify them, detect the language and draft replies in the commenter's language. Health questions are held back for the doctor.
Care team dashboard
Every comment and post across Facebook, YouTube and X in one place, with search, filters, AI-suggested replies, editing, escalation and CSV export. Admins and team members see different columns.
Deployment
Self-hosted n8n on a Bengaluru-region VPS, managed MySQL, Qdrant and Firebase phone authentication.
Outcomes
- comments processed and logged
- 1,900+
- comments processed and logged
- languages detected and replied in
- 2
- languages detected and replied in
- answers to medical questions, by design
- Doctor-only
- answers to medical questions, by design
Stack
- n8n
- GPT-4o
- Qdrant
- MySQL
- Firebase Auth
- Meta Graph API
- YouTube Data API
- Docker
- Caddy