Aryaman Garg

Queue Cure

The Waiting Room Nobody Fixed Go to any government hospital or small clinic in India. Not the big corporate one with the glass doors and the token machine that sometimes works. The real one — the one your parents go to, the one in your mohalla, the one where the doctor actually knows your name. You already know what you'll find there. Someone's written token numbers on torn paper slips. The receptionist is simultaneously talking to a patient, answering the phone, and trying to remember whose turn it actually is. And the waiting room — the waiting room is full of people doing the same silent c Process Honestly, I didn't sit down with a plan. I started by just thinking about that one image — a receptionist at a busy clinic, juggling three conversations at once, managing an entire queue from memory, and somehow holding it all together through sheer habit. That image bothered me enough that I opened my editor and started writing a server before I even had a proper idea of what the frontend would look like. The first thing I built was the queue state on the server — just an array, a current patient, and a broadcast — because I figured if I got that right, the screens would almost write themselves. They didn't, obviously. I spent more time on the "Call Next" button than on anything else, because I kept imagining what happens when a tired receptionist clicks it twice by accident, or when the Results The most honest result is this: both screens stay in sync. Every single time. Click "Call Next" on the receptionist screen and the waiting room board updates in under a second — no refresh, no lag, no moment where one screen says 42 and the other still says 41. That was the core promise of the project and it holds. The wait time estimates actually get more accurate as the day goes on. The first patient of the day gets an estimate based on whatever the receptionist typed in. Reflection The first thing I'd change is the data storage. Right now everything lives in memory on the server — which means if the server restarts, the entire queue disappears. During a hackathon that's fine. In a real clinic in the middle of a busy morning, that's a disaster. I'd move the queue state to Redis from day one. Not because it's complicated — it isn't — but because I spent time during this project knowing that gap existed and working around it mentally instead of just fixing it. That's the wrong order. Persistence should've been the foundation, not an afterthought I planned to add later. The