Gulyak.
All work
AI2024

MediTrack AI

Cut patient intake time by 60% with an AI-powered triage assistant.

8 weeks3 engineersClient: Private Medical Clinic
MediTrack AI

The Challenge

The clinic was processing 200+ daily patient inquiries via phone. Receptionists were overwhelmed — scheduling one appointment took 8–12 minutes. After-hours requests went entirely unanswered, causing lost bookings and frustrated patients. The team needed relief without hiring more staff.

Our Approach

We built a conversational AI triage assistant on top of GPT-4, tightly integrated with the clinic's existing scheduling API. The bot handles symptom intake, appointment booking, prescription renewal requests, and a full FAQ — 24 hours a day. Sensitive cases are escalated to a human automatically. A lightweight Next.js admin panel gives staff real-time visibility into conversation logs, unresolved threads, and booking volume.

  • GPT-4 fine-tuned prompt chain for medical triage — conservative escalation thresholds
  • FastAPI backend with async task queue for instant webhook responses
  • Bidirectional sync with the clinic's scheduling system via REST + polling fallback
  • HIPAA-aligned data handling: no PII stored in LLM context, all logs encrypted at rest
  • Admin dashboard with live conversation feed, resolution metrics, and export
Within two weeks of launch our receptionists stopped dreading Monday mornings. The bot handles the flood so they can focus on patients who actually need them.

Head of Operations, MediTrack Clinic

Results

60%
Reduction in intake time
24/7
Availability (was 9am–5pm)
More daily appointments handled
4.7★
Patient satisfaction score

Tech stack

Next.jsGPT-4FastAPIPostgreSQL

Have a similar project?

We'd love to hear about it. Free consultation, no strings attached.