100% On-Premise
Data stays in your plant. No cloud, no data leakage, no dependency on hyperscalers.
CURRENTLY IN PILOT
MINERVA
Edge AI for production.
Understand machine data — without the cloud.
Compute · Processing
Node board
TinyML on-device.
01 What is MINERVA
The local AI operator for machine data.
Sensor data is analyzed directly at the machine. In real time, without the data ever leaving the plant.
Modular hardware
Three-tier architecture (Probe / Node / Hub). Plug-and-play, industry-agnostic, retrofit-ready.
Energy-efficient
Edge inference uses 10–100× less energy than cloud processing. Built-in energy metering for CSRD.
Currently in pilot. Early adopters: HAHN Automation / Fissler.
02 The problem
Cloud IoT doesn't work on the shop floor.
Three reasons why today's IoT stacks fail in mid-sized manufacturing.
Data leakage & compliance
Machine data is the crown jewels of any production. With cloud IoT it ends up on someone else's servers, often in third countries. CLOUD Act, NIS2, and the EU AI Act turn this into a compliance trap.
Latency & availability
Predictive maintenance needs response times under 100 ms. A cloud round-trip makes that impossible. Any network outage stops the analysis — and the value with it.
Scaling cost
Cloud inference is priced per data point. With 1,000 sensors × 100 Hz, that quickly hits six figures per year — and the bill grows with every new machine.
MINERVA solves all three — through a different architectural principle.
03 The architecture
Three hardware tiers. One data philosophy.
Each tier scales independently. All data stays inside the plant.
At the machine
Reads diverse sensors and transmits the data via CAN bus to the Node. Multiple variants with different sensor packages available.
Near the machine
Collects sensor values from multiple Probes, processes them, and runs smaller AI models locally. Sends aggregated data to the Hub.
On-premise in the plant
Aggregates preprocessed results from all Nodes and runs the heavier AI models. Orchestrates the entire system.
04 Topology & scale
Hierarchical. Scalable. Robust.
One Hub orchestrates any number of Nodes, each Node reads any number of Probes.
WiFi wireless CAN bus (wired)
HUB one orchestrator
NODES one per machine
PROBES several per Node
Example: 1 Hub · 3 Nodes · 9 Probes
05 Hardware stack
Three module families. Freely combinable.
All boards: in-house development.
Processing Compute
Node board
STM32N6: low-power ML on-device
Hub board
Raspberry CM5 or Jetson Nano: headroom for large ML models
Connectivity Interface
IO-Link module
Bridges IO-Link sensors to the Node via CAN bus
SPAM module
Sensor & Peripheral Access Mesh, multi-probe bus
Measurement Sensor
Probe board
STM32H5: accelerometer, magnetometer, temperature
Energy board
KEY ASSET FOR CSRD REPORTINGCurrent, voltage and power measurement in real time
06 Key features
What sets MINERVA technically apart.
Local AI inference
Powerful carrier boards (CM5 / Jetson Nano) run ML models directly inside the plant — from tiny TinyML on the Node to multi-layer models on the Hub.
Energy-efficient MCUs
STM32 family from ST Electronics on Node and Probe boards. Sensor capture and inference in the milliwatt range — right at the machine.
Industrial interfaces
CAN bus for robust Probe–Node communication. IO-Link for standard industrial sensors. WiFi for Node–Hub. All Industry 4.0-ready.
Modular expansion
New sensors? Add a new interface module. More compute? Upgrade the Hub to Jetson. No vendor lock-in — open standards throughout.
07 Software · Live monitoring
Real-time view of every machine, every sensor.
A modern web front-end — multi-master, multi-site, without cloud dependency.
Live sensor values
Temperature, vibration, current — real-time updates straight from the edge hub.
Threshold alerts
Automatic warnings before the equipment takes damage.
Historical trends
Time-series per sensor — the foundation for predictive maintenance.
Multi-master overview
Multiple plants, multiple sites — all in one interface.
08 Software · Configuration
Plug-and-play, even for non-specialists.
Scan sensors, drag onto the plant photo, done. No IT specialist needed to set it up.
01
Scan the Node
New hardware is detected automatically — no manual IP configuration.
02
Place the Node
Drag-and-drop onto a photo of the plant — visual mapping without CAD.
03
Configure
Thresholds, sample rates, alarms — all directly in the web UI.
A modern front-end for the plant crew — not an engineering tool.
09 Energy & CSRD
MINERVA turns energy data into leverage.
What cloud solutions cannot do — and why it matters for CSRD reporting.
Energy board
Current, voltage and power measurement at the machine level, deployed as a Probe board directly at the consumer.
CSRD compliance without estimates
Direct measurement instead of guesses. Machine-level energy data as evidence in sustainability reporting.
Detect & smooth peak loads
Live load profiles per machine identify consumption peaks. Cuts grid fees and reduces energy costs.
Predictive maintenance + energy
Rising consumption is often the first sign of mechanical wear. One sensor stack, two use cases.
Industrial AI as an impact lever — not as another energy hog.
PILOT PROGRAM Pilot is running.
Pilot is running.
Early adopters welcome.
Running production lines and want to understand machine data in real time — without handing it to the cloud? We're looking for pilot partners to shape the next generation of industrial edge AI with us.