A wave of fine dust drifts through the dark, a green thread glowing at its core. In front of it, a notebook shows the MINERVA overview of readings.

Machine AI · Design partner program

MINERVA Get more out of your machines.

On the shop floor, without the cloud.

Become a design partner

Your machines are talking. MINERVA listens, measures right at the machine and notices when something changes. On the shop floor. Without the cloud.

At the machine

See the standstill. Before it happens.

MINERVA measures right at the machine. At the bearing, the spindle, the drive.

See the Probe? It sits right on the machine.

The Probe. Measures where it happens.

Vibration, temperature and magnetic field. On a board of its own.

The system

Three tiers. No black box.

Every board is an in-house design. We open them up for you, down to the chip.

Exploded view of the Probe board: the case opens, the board with its STM32H5 and the two M12 connectors floats free

The Probe · At the machine

Measures where it happens.

At the bearing, the spindle, the drive: the Probe sits right at the machine and reads the sensors.

Exploded view of the Node: lid and case move apart, revealing the board with its STM32N6, ESP32-C6 and expansion headers

The Node · Near the machine

Computes where it's measured.

The Node bundles up to three Probe chains, processes the values locally, and sends the results to the Hub wirelessly.

Exploded view of the Hub: the vented lid lifts off and the compute module rises from its carrier board with USB-C and M12, with the PoE board below

The Hub · On-premise in the plant

Orchestrates the whole plant.

The Hub collects the results from all Nodes and runs the AI models that are too large for the Node. This is where the measurement meets the knowledge in your documents, on-premise, inside the plant.

One Hub orchestrates any number of Nodes, each Node up to three Probe chains. Probes chain together on the CAN FD bus, and the Nodes reach the Hub wirelessly.

Example: 1 Hub · 3 Nodes · 9 Probes

Up close

Down to the detail. Our own hardware.

STM32N6 with AI accelerator.

TinyML runs right on the Node.

Bolted aluminum enclosure.

With the IGNAITE logo on the lid.

M12 for the shop floor.

CAN FD on Probe and SPAM, Gigabit Ethernet on the Hub.

Early warning

Bearing 1 gets louder. MINERVA notices.

1314151617181920 Warning above 0.12 g Alarm above 0.25 g Bearing 1Bearings 2 to 4 20 July, 07:10 1314151617181920 Warning above 0.12 g Alarm above 0.25 g Bearing 1Bearings 2 to 4 20 July, 07:10

13 July. Four bearings on line 3, all in the normal band.

From 17 July, bearing 1 climbs. MINERVA flags 44 abnormal readings.

20 July, 07:10. The vibration monitor shuts the line down.

RMS per bearing · 24-hour average · 13 to 20 July. Levels from the maintenance instruction. Example from a demo, not live operation.

The limit sits in a maintenance instruction, the last bearing service in the maintenance system, the similar case from March in someone's head. MINERVA measures right at the machine and brings the measurement together with all of it, on site, on your own hardware.

One memory

MINERVA listens. And LOKI answers.

One question from the early shift. The answer names its four sources.

Early shift, 07:14

Why is line 3 down?

The vibration monitor shut the line down at 07:10. The progression at bearing 1 is an outer race defect: remove, inspect, document the finding. Spare bearing Rexnord ZA-2115, replace as a set only.

  • 1Maintenance instruction WA-L3-014 · p. 1 to 2
  • 2Test rig bearing documentation · p. 2
  • 3Time series vibration_rms · bearings 1 to 4
  • 4Event log line 3 · 20 July, 07:10

What I cannot tell you

Confirming this needs a frequency analysis of the raw signal. In normal operation only condensed readings are archived; the raw signal for this bearing sits on the probe and is available through the API.

One question, both worlds: measurements from MINERVA, knowledge from the documents, one answer with twelve sources.

Example from a demo. The plant and documents were set up for it.

Shared memory

On the shop floor

What needs doing on line 3?

MINERVA. The drive shut off at 14:20 because of bearing 2. Vibration had been climbing since this morning, the same pattern as six weeks ago. Inspection step in manual 4.3, and the spare bearing is on the shelf. 3 sources used

In the office

How is production running right now?

LOKI. Three of four lines are running; line 3 has been down since 14:20. Utilization today is 82 %, with two unplanned stops, both on line 3. The weekly trend stays steady. 2 sources used

Machine data and company knowledge share one memory. A question from either side reaches all of it. Each also works on its own.

Effortless

Plug-and-play. Even for non-specialists.

Scan the Node, drag it onto the plant photo, done. The plant crew makes changes themselves. No IT project.

“Scan QR code” dialog in MINERVA: the camera captures a Node’s label with its QR code and ID, with the Node placement and the plant photo behind it.

Scan the Node. MINERVA detects new hardware automatically. No manual IP configuration.

Node placement in MINERVA: the scanned Nodes with their Probes on the left, the photo of hall 1, line 3 on the right with Probes placed at robot R2, station T07 and conveyor M1.

Place the Node. Drag the Node onto a photo of the plant and map it visually, without CAD.

“Configure probe” dialog for robot R2: the Probe offers its own settings, such as sample rate, alarm on threshold violation, thresholds for motor temperature and vibration, and the calibration offset.

Configure. Set thresholds, sample rates and alarms, right in the web UI.

MINERVA system overview: all Nodes with their Probes on the left, the live values of one Node for motor temperature, current and vibration with history on the right.

Live values and history. A modern front-end for the plant crew. Not an engineering tool.

Example. Hall, line and names are made up.

The lineup

Seven boards. One kit.

Compute, interface, sensing: every board is an in-house design, freely combinable.

Node board: STM32N6: low-power ML on-device

Compute · Processing

Node board

TinyML on-device.

Hub board: Raspberry Pi CM5: runs the AI on the CPU, passively cooled

Compute · Processing

Hub board

The AI orchestrator.

Hub with GPU: NVIDIA Jetson Orin Nano with GPU for larger AI models

Compute · Processing

Hub with GPU

For larger AI models.

IO-Link module: Bridges IO-Link sensors to the Node via CAN FD

Interface · Connectivity

IO-Link module

IO-Link, now on the CAN FD bus.

SPAM module: Three M12 ports, each with its own CAN FD bus and switched power

Interface · Connectivity

SPAM module

Bus and power for the Node.

Probe board: STM32H5: vibration up to 6 kHz, acceleration and angular rate, magnetic field, temperature, humidity, air pressure, gas

Sensor · Measurement

Probe board

Vibration, temperature, magnetics.

Energy board: Current, voltage and power measurement in real time

Sensor · Measurement

Energy board

Current, voltage, power. In real time.

KEY ASSET FOR CSRD REPORTING

All boards developed in-house, no white-label, no reseller.

Energy & CSRD

Turns energy data into leverage.

Cloud solutions can't do this. And CSRD reporting depends on it.

The MINERVA energy board on a clear background: a circuit board with current and voltage measurement channels, terminal blocks and a microcontroller.
Energy board. Current, voltage and power measurement at the machine level, deployed as a Probe board directly at the consumer.

10 to 100×less energy than cloud processing.

CSRD compliance without estimates

Back your sustainability reporting with machine-level energy data. Measured directly instead of estimated.

Detect & smooth peak loads

Live load profiles show each machine's consumption peaks. So you cut grid fees and energy costs.

Predictive maintenance + energy

Rising consumption is often the first sign of mechanical wear. One sensor stack, two use cases.

So industrial AI saves energy instead of using up more. And you query the load profiles in plain sentences.

No lock-in

Your system. Your data. Your pace.

MINERVA runs fully offline behind the plant firewall. No cloud dependency, no data leakage.

The MINERVA Hub in its closed aluminum enclosure, with flame engraving, USB-C and an M12 connector.

No internet on the shop floor?It runs.

Your MES wants the data?

It gets them via API, MQTT, or straight from the database. No third-party cloud, no paywall: you connect your systems, not the other way around.

IO-Link sensors already in the machine?

They stay. The IO-Link module brings them onto CAN FD and into the chain.

A sensor no longer fits?

Probe off, new one on, detected automatically. The measuring point grows with you.

Models outgrowing the Hub?

Upgrade the Hub board to a Jetson Orin Nano. Everything else stays as it is.

Integrating for clients?

Plug in a new interface module and the Node detects it automatically. Open standards: CAN FD, IO-Link, WiFi. Not a black box.

Every interface and component is listed in the tech specs. Tech Specs

Platform

Four tools. One platform.

What MINERVA measures, LOKI explains in full sentences, with maintenance instructions and history.

The design partner program is running. Limited places for 2026.

If you run production lines and want to understand machine data in real time, without handing it to the cloud, come in as a design partner. You get MINERVA on the shop floor early. Your requirements go straight into development. And you lock in first-series terms.

For manufacturers, automation specialists, integrators, and system houses.

Still have questions?

Read the FAQ