Pilot program customers

Kortex Dynamics deploys autonomous inspection robots in semiconductor fab environments. Battery life is critical: robots run 16+ hour shifts between charges. Their previous TFLite Micro anomaly detection stack drew 240 mW active, limiting shift length.

"We evaluated three edge inference approaches before Neurmorph. The power reduction on the NT3000 was the first result that actually moved the needle on shift duration. The migration from TFLite took two days."

— VP Engineering, Kortex Dynamics

Halvane Systems develops autonomous sensing payloads for small UAS platforms operating in contested environments. Radio silence protocols require completely battery-powered inference with no external communication during mission. Previous GPU-based approach was power-prohibitive.

"Sub-40 mW active inference was what we needed to make the payload viable on standard UAS battery budgets. Neurmorph's static memory layout was critical for certifiability — no dynamic allocation means predictable behavior under fault conditions."

— Chief Systems Architect, Halvane Systems

Virelux Sensing deploys multi-modal environmental sensors in remote industrial sites. Sensors run on harvested energy — solar + vibration — with no reliable grid connection. Their previous inference stack required a Cortex-M7 running at full clock, consuming more power than the harvester could sustain continuously.

"The event-driven runtime model was a better architectural fit for our sparse sensor data than any always-on approach. The NT3000's dormancy power at 3.8 mW is below what our harvester can sustain indefinitely."

— Head of Hardware, Virelux Sensing