Safety monitoring · Federated AI · Built in Estonia
Watch over someone continuously, without a single byte of their data leaving their pocket.
HeliTrust turns a phone someone already owns into a round-the-clock safety monitor: falls, stillness, wandering, fatigue, shock. Unifying Contextual Weights evaluate motion telemetry on-device so raw sensor readings never leave the hardware.
1s telemetry · 50Hz sampling · Unifying Contextual Weights · 6 domains, one engine
The core innovation
GDPR compliance that isn't a promise in a contract; it's a fact about the architecture.
Every other approach to this problem sends raw health or location data to a cloud server and applies privacy law after the fact. HeliTrust removes raw data from the equation entirely using Unifying Contextual Weights; the server never receives a sensor reading, only anonymous weight updates from the device's baseline model.
Read the full architecture walkthrough →One engine, six calibrations
The same telemetry pipeline watches for six very different things.
A resident, a lone worker, a child, an athlete, and a piece of machinery all produce a motion signature. What counts as an anomaly differs for each, so HeliTrust calibrates Unifying Contextual Weights across domains while maintaining the exact same privacy-first pipeline.
HeliTrust Care
ELDERCARE
WHAT IT WATCHES FOR
- • Falls, the instant they happen
- • Wandering & safe-zone breaches
- • Circadian rhythm deviation
The problem: A facility with 60 residents and 4 night staff isn't failing by checking hourly; hourly is near the physical limit of that staffing ratio. Gradual decline can go unnoticed for weeks because no one is positioned to compare today against three weeks ago, across dozens of residents at once.
How HeliTrust solves it: Fall-magnitude detection flags a fall the instant it happens. A background stillness sweep catches the case where someone goes quiet without an obvious fall. An EWMA-based personal baseline makes gradual decline visible; '10% less active than last month' is invisible to spot checks but exactly what a running statistical baseline catches.
What each person actually sees
Three very different screens, one underlying record.
The person being monitored, the caregiver watching over them, and the administrator proving compliance never need the same view, so HeliTrust doesn't give them one.
-
A plain-language status, not a number
Normal, elevated, or alert; never a raw risk score or sensor reading on someone else's screen.
-
SOS, always on screen
One tap sends an immediate alert to caregivers with the person's ID and last known location, even mid-shift or mid-pause.
-
A ten-second daily check-in
A quick mood, pain, and sleep self-report from the phone; caregivers see who has checked in today without anyone having to ask.
-
Session and check-in tracking
For shift-based and lone-worker roles, a missed check-in is caught automatically instead of depending on someone remembering to look.
-
An honest pause control
A short, auto-expiring pause for privacy; SOS still always works, and every pause is logged, so "was monitoring off" is always answerable.
-
Your own one-time setup code
Each device is enrolled with a private code issued just for it, not a password shared across a whole facility or family.
-
A family portal that reassures, not exposes
A simple "all good" or "check needed" view, no login required, no raw sensor data ever shown to family.
What makes this different
Measured against how this category is usually built.
Not a comparison of brands; a comparison of two architectures. One requires proprietary hardware and treats privacy as a contract; the other doesn't.
| Capability | HeliTrust | The typical approach in this category |
|---|---|---|
| Hardware | Any Android phone the person already owns | A €100–200 proprietary pendant, purchased per person |
| Data privacy | Technically enforced; raw data never leaves the device | Contractually promised; raw data reaches a cloud server |
| EU AI Act readiness | Article 22 human oversight and a 10-year audit log, built in | Generally not yet assessed against the 2027 deadline |
| Platform breadth | One engine, six safety-critical domains | A single-purpose product, one vertical only |
| Who it fits | A single family through to a multi-site organization, each with isolated data | Usually sized for one or the other, not both |
Why this matters now
The staffing math doesn't work. Continuous monitoring is the part that doesn't need more headcount.
€25B
Annual cost of falls to European healthcare systems
63%
Of assisted living facilities report a staff shortage
1 in 3
Europeans will be 60 or older by 2050
Partnership & investment
Ready for a pilot in eldercare, workplace safety, or anywhere continuous monitoring replaces a scheduled check.
A live demonstration, a technical architecture walkthrough under NDA, or a clinical validation study design, available on request, across any of the six domains.