The bell needs hands and words
A call button assumes the person can reach it, press it and explain. Many residents can do none of these.
Winner · Yozmim Chaim hackathon final, Adi Negev 2026
When a resident can’t call for help, the room calls for them.
Adi Monitor turns a resident’s breathing, vital signs, movement and voice into their call bell. Simple sensors watch for danger, and the nurse’s phone rings with what happened, to whom, where and why, escalating until someone answers.
The problem
Many residents cannot press a call bell.
In rehabilitation villages and care homes, people with severe physical or cognitive disabilities depend on someone noticing. When they choke, stop breathing well, fall or are in distress, nobody is called unless staff happen to be in the room.
Nights are long
Rounds are spaced and staff are few. The minutes between rounds are exactly when breathing problems are missed.
Noisy monitors get ignored
Hospital monitors raise so many false alarms that staff learn to tune them out. A new alarm must earn trust.
See it work
A working prototype, end to end.
Real backend, real nurse app, simulated sensors and the laptop’s real microphone: a new sensor plugged in with zero code, a resident who stops breathing, an escalation nobody answered, a sensor pulled out, and a rule changed live.
How it works
Sensors in the room, a server in the building, a phone in every nurse’s pocket.
Sensors
A breathing radar by the bed, an oximeter and thermometer on the resident, a bed-exit pad, a room microphone and the call bells already there.
Room computer
One small computer per room collects the sensors and sends only events and scores, never recordings.
Local server
Decides: rules per measurement, each resident’s own normal values, two independent signals before some alerts, a record of everything.
Nurse station and phones
What happened, to whom, where and why. Phones ring; if nobody answers, it escalates.
Everything runs inside the building, on its own network. Nothing goes to the cloud.
Nobody answers? It escalates.
An alert never just sits on a screen.
- 0 s
Every nurse’s phone
A critical alert rings the phones of the nurses on the ward.
- 30 s
The charge nurse too
Not acknowledged in time: the shift lead is paged as well.
- every 2 min
Reminders
Until someone acknowledges, both are reminded again.
- always
On the record
Every page is logged with how many phones it reached, including zero.
Why it’s different
Built so a nurse can trust it.
Missing data is never zero
A silent sensor raises “sensor offline”, never a false “not breathing”.
Fewer false alarms
Two independent signals must agree before some alerts, and each resident is compared with their own usual values.
Never fails silently
Sensors check in every 5 seconds. A screen that lost its connection says “not live” instead of looking calm.
A new sensor needs no code
Plug it in, approve it to a room, a bed or a resident. That’s it.
Private by design
Sound is analysed in the room; only a score leaves it. Nothing is recorded, nothing goes to the cloud.
Every decision on record
Approvals, alerts, pages and rule changes are journaled: who knew what, and when.
The screens
Clear at a glance, in Hebrew and English.
Where we are
1stplace, hackathon final
≈550automated tests on every change
3surfaces: station, phone, server
2languages, right to left included
Roadmap
From prototype to a real room.
Working prototype
- Server, nurse station and Android app
- Escalation to phones, on the record
- Simulated ward and a real microphone
Pilot room
- A room computer with a breathing radar and Bluetooth oximeter
- Clinicians set every threshold
- Loud alarms on nurses’ phones
A product to trust
- Accounts, roles and encrypted links
- Nurse-call system integration
- Sound recognition trained with consent
Learning and scale
- Rules tuned by nurses’ real/false feedback
- Medical-record integration (FHIR)
- More wards and sites, same setup
Let’s talk
We are looking for a pilot site, clinical partners and early supporters.