Magnetic spiking neurons, built from a domain wall and a trap.
Like a soap bubble, a magnetic bubble is charged and triggered by each impulse of the raw signal, just like a biological neuron. Sensitive to the temporal encoding of information, our neurons detect the timing of complex events, not only their frequency components.



Inspired by human hearing.
The MAG.NET project, funded by the EIC Transition programme from 2023, converts signals into spikes in the way the mammalian ear does. The physics behind it was discovered at Spintec-CNRS: a domain wall escaping a geometric trap behaves in many ways like a spiking neuron.
- Accuracy
- Deep-learning levelWith 10x less training data and compute.
- Fine-tuning
- None per machineVersatile like hearing, from one asset to a whole plant.
- Sensitivity
- Temporal and frequentialNot limited to FFT spectra.
- Versions
- Hardware and softwareProprietary ML algorithms built in tandem with the neurons.
An intuitive dashboard, one report per anomaly.
Every anomaly gets its own signature, so complex signals stay readable over time. One click opens a root-cause analysis, and the report recommends actions before damage sets in.
How does Magnetic Spiking Intelligence detect anomalies?
The edge module samples raw analog signals at 10 kHz or more, converts them into timing-rich fingerprints and compares each new fingerprint to a self-learning baseline in real time. Any drift outside the normal envelope is flagged within milliseconds.
Can it monitor several signal types at once?
Yes. The platform accepts any analog signal at 10 kHz or more: current, vibration, magnetic field, acoustic, optical, temperature and more, with no re-engineering.
How long does the baseline period take?
Typically 3 hours to 5 days, depending on cycle time and process variability.
Can I move the device to another asset?
Yes. Relocate the module and trigger a one-click re-learn to build a new baseline on the new equipment.
Can the system run fully on-prem?
All analytics run on the edge device. For storage and dashboards you choose a sealed server inside your network, your own cloud, or the Golana managed cloud, and you can switch later.
How does it connect to SCADA, MES or CMMS?
Golana develops interfaces with the platform of your choice.