Edge AI · Live demo

Predict machine failures before they happen.

Train a model on your data in 30 seconds. Then watch it spot faults on an ESP32.

Try it now ↓ See the ESP32

The finished model runs offline, no cloud.

§01Try it yourself

Train a model in 30 seconds.

Upload a CSV or pick a sample dataset. No signup.

Vivi engine · training
Upload CSV
or try a sample dataset
Training in progress...
Training Results
Accuracy
Parameters
Inference Time
Samples
Features
Classes
Sign up to download the compiled binary
Create Free Account →
ERR
§02On ESP32

The same model, on an 8-euro chip.

97 KB of pure C. No internet, no cloud.

Predictive maintenance · live simulation ESP32 — 240 MHz — 512 KB RAM
100%
Fault detection accuracy
97KB
Model size in flash
< 1ms
Inference time
€8
Total hardware cost
§03Fault Detection

Four faults. One chip.

8 accelerometers, 4 motor states, in real time.

01

Normal

Regular vibration. Motor within spec.

02

Rotor Imbalance

Strong 1x component. Balance within 2 weeks.

03

Bearing Wear

High-frequency impulses. Replace before it fails.

04

Misalignment

Strong 2x and 3x harmonics. Realign the coupling.

How Edge AI works →

Protect your machines.

From CSV to compiled C in minutes. Start free.

Start Free → Talk to Sales