Monitor
Anomalies detected in raw data with the accuracy of deep learning, on any signal from any machine.
Golana Computing turns raw vibration, current and acoustic signals into anomaly fingerprints within milliseconds, explains the physical cause and proposes the action to take. A Spintec-CNRS spin-off from Grenoble.
Each Golana neuron is a magnetic domain. Every impulse pushes its wall against a geometric trap, and it only escapes when the timing adds up, just like a biological neuron. The result is deep-learning accuracy with 10x less data and compute.
The edge module samples any analog signal at 10 kHz or more and compares each new fingerprint with a self-learning baseline. When patterns drift, the AI pinpoints the root cause and writes a report with recommended interventions.
No machine-specific fine-tuning. Move the module to another asset and trigger a one-click re-learn.
Ask for a demo
Physics, magnetism, microelectronics, software and machine learning under one roof in Grenoble.
Meet the teamAn all-in-one system, from data acquisition to the intervention list. Hardware is retrofitted on existing equipment or integrated into new machines during development. Software runs as a Maintenance or Process SaaS.
Anomalies detected in raw data with the accuracy of deep learning, on any signal from any machine.
A signal analysis toolbox identifies and explains each anomaly.
AI reasoning about the physical phenomenon behind the anomaly.
Interventions ranked using machine specs, history and costs.
Because all raw data is kept, the colour code can be replayed over long periods to find recurrent problems. Each tile is a fingerprint of healthy operation, so the slightest drift stands out.
Live partner deployments in sequential and continuous operation, from medium to extreme process variability.
Sequential operation, multiple processes with extreme variability. Ageing of laser components.

Surveillance of multiple components, such as the mould, to avoid catastrophic failure.

Continuous operation, a single process with large variability. Preventing short-circuits in the battery.
Continuous operation, multiple processes with medium variability.

Founded in 2023 to industrialise a breakthrough in the design and fabrication of bio-mimicking magnetic neurons, developed for years in the SOFT project. The discovery behind it: a domain wall escaping a geometric trap behaves in many ways like a spiking biological neuron.
How the neurons workFixed-price hardware or rental, plus a subscription to the dashboard. Retrofit in hours, baseline in days.
Ask for a demoIntegrate the monitoring into new equipment during development and fabrication, and ship machines that explain themselves.
Discuss integrationBacked by the EIC Transition programme since 2023 and selected for the EIC Accelerator. Hardware and software versions of the neurons, protected by patents.
Talk to the founders
MAG.NET, Magnetic neural Network for predictive maintenance, EIC Transition 2023 to 2026. Follow-up EIC Accelerator project with trials at seven industrial partners in Europe.