Carsten Sinke
← All work
Momo Medical · 2025 – now

Every resident, on one screen

Sensors under the mattress tell nurses who is getting up, before anyone falls. I work across the app nurses carry and the services behind it.

  • Flutter
  • Node.js
  • TypeScript
  • PostgreSQL
  • AWS
Tap a room, or wait for a help request

Three in the morning. A nursing home at night. A handful of nurses, a long corridor of rooms, and residents who should not get out of bed on their own. A fall at that hour can change the rest of someone's life.

Momo puts a sensor under each mattress. It knows who is asleep, who is restless and who is about to stand up, and tells the nurse before it happens.

The app goes wherever the nurse goes: phone, tablet or the screen at the nurses' station.

Where I come in. I joined Momo in the summer of 2025 to work on the app nurses carry all night, and on the services behind it. It runs on phones, tablets and the web, for care organisations that each work a little differently.

The bar is simple to say and hard to meet: it must never be misread, and it must never get in the way. Three pieces of that work are below.

A screen that shouts all night stops being heard.
Speak-listen

Talk to the room from your pocket

A button on a wrist. A resident presses the help button on their wrist. Somewhere down the corridor, a nurse's phone lights up: Room 12 is asking for help.

Before, the only way to find out what was wrong was to walk there. Maybe it was a fall. Maybe it was a glass of water. With Lizzy, the speaker beside the bed, the nurse can simply ask.

Call the room, then hang up

Building the call. I wrote most of the Flutter code that makes that call: setting up the WebRTC connection to Lizzy, the call screen, the volume a resident can actually hear, speaker and mute, and a hang-up that really ends the call on the device too.

It had to work on Android, on iPhone and in the browser at the nurses' station, where I used Web Audio.

The day the build broke. Then Momo added video calls with a second provider, 100ms. The moment both libraries were in the app, Android refused to build. Each one shipped its own copy of the same WebRTC classes, and the two could not live side by side.

We could wait for the vendor, drop a feature, or dig in.

One screen for every call. With both libraries in, I rebuilt calling around one strategy: Lizzy and video calls now share one call screen and one call log. I added logs with enough context to trace a failed call the same day, and tests for every permission a phone can refuse.

I forked their SDK, moved it onto its own WebRTC build, and we shipped.
1M+calls a year between nurses and residents
3platforms: iOS, Android and the web
Sign-in

Signing in, on a tablet the whole ward shares

One tablet, a whole ward. At shift change a tablet goes from one nurse to the next. It stays signed in for the whole ward, because nobody has time to log in at a bedside. But some things, like a resident's care details, should know exactly who is holding it.

And every care organisation has its own accounts, its own identity provider and its own rules.

Moving the front door. I moved Momo's whole sign-in to Auth0: the Flutter app, the backend and the authentication service. Care homes now sign in with their own accounts through single sign-on, and the groups people belong to come along from their organisation.

There is no quiet night to switch over in a care home.

Switching while everyone is working. A nursing home never closes, so there was no maintenance window. Old and new tokens ran side by side, and I measured per device which one it still used. Only when nobody needed the old way did it go.

Sign in, then open care details

Asking who you are, only when it matters. On a shared account, opening care details now asks the nurse to confirm it is them, with a link sent to their own email that works for fifteen minutes. One place in the app asks for it and then carries on with what she was doing, so no screen has to remember to.

I counted every step, links sent, opened and expired, so we could see where people got stuck instead of guessing.

Becoming the Auth0 person. Along the way I learned Auth0 inside out, and wrote it down so the team does not depend on me for it.

300+nursing homes whose staff sign in through it
SSOwith each care home's own accounts
MFAon tablets the whole ward shares
Product

Screens the server can change

More and more integrations. Momo connects to more and more systems, and each one brings its own settings for rooms, residents and devices. Every new integration used to mean new screens in the app: a build, a store review, and then waiting until every phone and tablet on every ward had updated.

What HEMA taught me. At HEMA the server draws the app's screens, and I learned there what that takes. I introduced the idea at Momo: the app knows a set of building blocks, and the server describes each screen with them.

It started with the settings for rooms, residents and devices. A new integration now adds its settings on the server, and every app shows them without an update.

A backend made for the app. Behind it grows a backend for the frontend. The business rules move out of the app and into that one place, so the app no longer has to know them, only how to draw what it is given. The team keeps building it out, one screen at a time.

Small things that matter at 3 AM. When a care home switches notifications off, an overdue turn used to show in the same alarm red as a real emergency. I changed it to a calm yellow, so red keeps its meaning.

Leaving it cleaner. I brought the app's static analysis down to zero findings and made the build refuse new ones. I set the design system next to the app and listed every place they still differed, so design and code can move together.

The app knows how to show things. The server knows what to show.
No releaseneeded to add an integration to the settings
0static analysis findings, enforced in CI
20,000+beds with a BedSense
5countries where Momo is live
25%fewer falls, measured by Vilans