Booligys reads swimmer posture in real time and recognizes the instinctive struggle that precedes submersion — the moment rescue is still possible, not after the person is already underwater.
In the United States, drowning remains one of the leading causes of accidental death in children — and even when it isn't fatal, the human and financial cost is out of proportion to almost any other kind of home accident.
Safety bodies like the U.S. Consumer Product Safety Commission recommend a "layers of protection" model — a physical barrier, an alarm, and adult supervision — and are explicit that no single layer is enough on its own. Traditional pool alarms react after the fact. Human supervision lapses for minutes at a time. Booligys is built to be the layer that watches continuously and reacts while there's still time to act.
Booligys pairs a surface camera with an NVIDIA Jetson Orin running computer-vision models locally — no cloud dependency for the critical decision — to recognize how a drowning person actually behaves, not just where they are.
Aquatic-rescue research documents a very specific physical pattern in the final 20 to 60 seconds before submersion: an upright body position, the head tilted back, arms pressing down and out instead of stroking, no forward movement, no ability to call out. Booligys recognizes this pattern from pose estimation on a surface camera and fires an alert within seconds — while the person is still visible, inside the real rescue window.
Stature classification from calibrated camera geometry — no facial recognition — identifies when a child is in the pool area with no adult nearby, covering unsupervised access, one of the leading causes of residential child drowning.
A surface camera streams live video to a local edge unit — no footage leaves the property.
A pose-estimation model tracks body position frame by frame, classifying behavior against known drowning patterns.
A confirmed pattern triggers a local siren and a push notification with the exact location — while rescue is still possible.
100% on-device processing. No raw video sent to the cloud, no facial recognition — the critical decision runs locally on the Jetson Orin, cutting latency and privacy friction by design.
Global drowning-detection AI market, 12.3% CAGR (2025–2034). The U.S. segment alone is worth $166.3M in 2024, growing 10.8% per year.
We're building toward a first validated pilot installation. If you'd like to learn more about the technology, the roadmap, or how to get involved, reach out directly.