Atcom Next (2016)
Phygital dining at Atcom Next 2016 — iBeacon coffee service, AR invitations, and a real-time burger kitchen powered by NodeJS and WebSockets.
Atcom · iOS · NodeJS · AR
- Atcom
- 2016
- Lead Architect / Developer
Atcom Next is Atcom S.A.’s annual Athens showcase of where digital experience is heading. In 2016 the theme was Phygital Dining — merging physical venue operations with apps, sensors, and real-time kitchen systems.
We were implementing AR and proximity experiences in 2015–2016, before ARKit/ARCore made that path common. The event was as much an engineering stress test as a demo day: hundreds of people, metal, radio noise, and no patience for brittle demos.
Coffee ordering — proximity-based service
Attendees got coffee as they approached the bar — no advance queue, no verbal order at the counter.
Physical invitations required the Atcom NEXT app beforehand, so we already had name, profile image, and (if they used LinkedIn login) job title. In the app’s “Coffee & Burger” section, guests set coffee preferences.
As someone entered the beacon zone, their face and profile appeared on a large front-of-house screen; the bartender saw face + preference on a dedicated display and could start the drink before the guest reached the counter.
Beacon technology and signal processing
iBeacons around the bench shared UUIDs and ran at high transmit power so the whole coffee area read as one zone. Stronger signal improved pickup — and made the hard problem sharper: when to fire enter/exit events.
We used received signal strength to decide “inside the zone” vs distance-from-source heuristics, tuned for ~600 people in the room with only a dozen waiting to be served at once.
Background location ran on iOS and Android. On iOS we also used Significant Location Change so the app could wake from a killed state when an attendee first approached the venue.
Events went to a NodeJS server, which streamed enter/exit updates to bartender and display screens over socket.io.

I built the event iOS app — registration, meal preferences, and proximity-driven barista notifications. It also included AR registration: point the camera at the physical invitation and a 3D model showed the attendee’s name. Pre-ARKit, that stack was OpenGL/C++ on camera sensor input.
Interactive ordering — Phygital Lunch
Windows touch tables ran a playful burger-ordering UI that posted HTTP events to the NodeJS backend. Orders linked to attendees via QR scan; the kitchen and grill boards updated in real time.
The grill screen showed counts by type and doneness; two cook screens ran one-order-at-a-time queues. HTTP + WebSockets kept state resilient under flaky venue Wi‑Fi. When food was ready, the attendee got a push notification.
On this track I owned the food-ordering backend (NodeJS + WebSockets).
Talk
Paris Stagkopoulos and I presented how the system was designed end to end:
Key learnings — ambient intelligence
Ambient intelligence fails in the gap between digital assumptions and physical reality. Metal, RF interference, dead spots, and unexpected crowd behavior can invalidate a lab-perfect design in minutes. Several full-day implementations were thrown away before something survived contact with a packed venue.