Edge AI · Industrial condition monitoring
Vaynova builds ultra-low-power edge AI for industrial machine-health monitoring. We run fault detection directly on the machine, in under a watt — no cloud round-trip, no latency, no raw data leaving the asset.
The problem
Industrial machines announce their own failures — in their vibrations — long before they break. The trouble is that the intelligence to hear those warnings sits in the cloud, held back by power budgets, bandwidth, and data-privacy limits. The sensors are already on the machines. The inference isn't.
The platform
A compact neural-inference engine for vibration-based fault detection. Built on an efficient depthwise-separable architecture and designed for sub-watt FPGA deployment at the sensor node, it delivers real-time condition monitoring without streaming raw data anywhere.
Inference runs at the sensor. Detection happens locally, in real time, with nothing raw sent off the asset.
An architecture engineered from the ground up to fit the power and memory budget of an edge node.
Sensing across multiple axes recovers early-stage faults that single-axis monitoring structurally misses.
Validation
Through a research-data collaboration with the AMRC (Advanced Manufacturing Research Centre, University of Sheffield), we've validated our models in software on real machine-tool and industrial-robot data. One finding stands out: multi-axis sensing recovers subtle faults that single-axis monitoring cannot detect.
Applications
Manufacturing, robotics, utilities, and logistics all depend on rotating and moving assets that fail expensively and without warning. EdgeSense brings predictive maintenance to any of them.
About
Vaynova is a founder-led, pre-seed semiconductor-IP company based in the UK. We design the low-power AI that lets industrial machines monitor their own health — making machine intelligence small, efficient, and trustworthy enough to live on every sensor.
A physicist-turned-engineer who built the EdgeSense platform from the ground up — from the signal-processing pipeline to the inference architecture and its FPGA implementation plan.
Contact
Working on condition monitoring, predictive maintenance, or edge AI — or want to know more about what we're building? Reach out directly.