One analog AI platform for always-on security across smart home, commercial property, fleet, and vehicle — all under 100µA. It listens for the events that matter and wakes the system only when something relevant happens.
Aspinity provides curated, pre-built detection models on the AML100 core so you can add proven capabilities without training from scratch. Available today:

Reliably detects the acoustic signature of breaking glass and triggers intrusion and security systems.

Recognizes standard T3 and T4 temporal alarm patterns acoustically, for smoke, CO, and other life-safety alerts. Works from up to 7m away.

Detects break-in, forced entry, and vandalism using acoustic and vibration signatures together for accurate, lower-false-alarm awareness.
All curated models run on the AML100-based MARC100 platform at microamp always-on power. Custom models can be developed for your application.
Traditional always-on sensors force a choice: accuracy or battery life. A DSP sensor sips power but false-alarms on pots, pans, barking dogs, and dropped items; a digital tinyML sensor is accurate but burns through batteries in one to three years. AnalogML delivers both accuracy and multi-year battery life.
By keeping sensing, processing, and decision-making entirely in the analog domain, the AML100 cuts always-on system power by more than 95% — waking the digital system only on a real event.
Today's solutions compromise accuracy to avoid draining a parked vehicle's battery. AnalogML monitors an unattended vehicle at under 50µA, essentially unnoticed by the battery.
Using sensor fusion across acoustic, vibration, and radar, AML100 recognizes the events that matter — touch, scratch, impact, glass break, forced entry — while ignoring the ones that trigger conventional accelerometers, like a neighboring alarm, a passing truck, or a runaway shopping cart.
Whether it's a consumer device or an automotive program, we'd like to understand your power and detection targets.