Queue Cure '26 — Real-Time Clinic Queue System

76% of India's 1.5 million clinics still run on paper token slips and shouted announcements. Patients wait 2-3 hours with zero visibility into when they'll be called. Receptionists juggle a physical register while manually tracking every patient. Doctors have no view into queue status until a complaint reaches them. Queue Cure '26 challenged builders to fix this with a digital queue where a receptionist's "Call Next" click updates the patient's screen instantly — no refresh, no shouting, no guessing. Process I scoped the must-haves from the brief: Receptionist Dashboard, Patient Dashboard, Add Patient, Call Next, Average Consultation Time, Real-Time Updates, Database Storage. I prototyped fast with a single React app, then rebuilt it as a Node/Express + Socket.IO server backed by SQLite, syncing across separate physical devices, not just browser tabs. I tested cross-device sync over WiFi, hit constraints like network client isolation, and solved it with a mobile hotspot fallback. For the AI Wait-Time Prediction bonus, I built a model that starts on the manual estimate and switches to a weighted average of real consultation durations once 5+ are recorded. I also added QR check-in, a WhatsApp deep-link, and browser push notifications via a service worker — each tested live on real devices. Results All 7 must-have requirements are fully working and tested: both dashboards, Add Patient, Call Next, Average Consultation Time, real-time sync, and SQLite storage. I verified real-time sync isn't just theoretical — I tested it live across a laptop and a separate phone on the same network, watching the patient screen update the instant Call Next was clicked. All 3 bonus features (AI prediction, QR check-in, analytics) work, plus two extras beyond the brief: WhatsApp deep-links and automatic browser notifications. Reflection I'd set up the real SQLite database and Socket.IO architecture from hour one instead of starting with a simpler BroadcastChannel prototype — the rebuild cost time I could've spent polishing the AI model further. I'd also test cross-device networking (WiFi client isolation, HTTPS requirements for notifications) on day one rather than discovering those constraints late. Given more time, I'd replace the WhatsApp deep-link with real Twilio SMS, and turn the reason-for-visit pricing into a configurable admin panel instead of a hardcoded lookup table.