Queue Cure '26 — Real-time Clinic Queue Management System

India has 1.5 million small clinics still running on paper tokens and shouting names. Patients wait 2-3 hours with zero visibility into when they'll be called. They can't leave, can't track progress, and have no way to know if they'll be seen soon or in 2 hours. Receptionists manage everything from memory. Doctors step out to check who's next. The entire system runs on guesswork — and it costs patients hours of their day, every single visit. Process We started by identifying what each stakeholder actually needs — not what's technically impressive. A patient needs one number and one sentence. A receptionist needs to add walk-ins fast. A doctor needs to call next without leaving the room. We chose Django Channels for WebSocket support so queue updates push to every device in under 50ms without polling. Token format went through one iteration — we initially used patient initials (AS-01) but switched to doctor letters (A-01, B-01) for privacy and instant room identification. For wait time we implemented EMA (α=0.3) instead of a fixed average — it adapts to each doctor's actual pace throughout the day. We built a full in-memory fallback so the app runs without MongoDB for demo purposes, then connected Atlas for production. Results Live app at democlinic.up.railway.app — patients check in with 3 taps, queue updates sync across all devices in ~50ms via WebSocket, wait time adapts dynamically using EMA per doctor. 4 doctor queues running simultaneously with separate logins, zero app install required for patients. Reflection I would add SMS notifications via Twilio from day one — the phone number is collected at check-in but never used. Many patients in small clinics are older and won't keep a browser tab open; an SMS when they're 2 away would make the system genuinely useful for them. I'd also build a proper admin panel earlier. Right now changing doctor fees or timings requires a direct database edit, which is fine for a hackathon but unusable for a real clinic owner. Finally, I underestimated how much time Windows-Linux environment differences would cost — git branch naming, line endings, static file s