How It Works

Hardware in your cooler.
Data on your phone

A compact sensor unit mounts to your storage ceiling. It runs computer vision models on-device, syncs with the Veratori dashboard, and sends your team alerts — no IT department required.

Setup

From unboxing to live in one day

No enterprise IT project. No lengthy onboarding. Just a sensor, a ceiling mount, and 24 hours.

01

Hardware installs in 30 minutes

A compact sensor unit mounts to your walk-in ceiling. No drilling into refrigeration, no network configuration required. Power it on and it's live.

02

Models calibrate overnight

The system spends its first 12–24 hours learning your specific inventory: labels, packaging, shelf positions, and containers unique to your operation.

03

Your team gets a dashboard

From day two onward, managers see a live inventory view, receive morning digests, and get instant alerts for anomalies — no training required.

Features

What Veratori does for your team

Three core capabilities that replace manual inventory from day one.

Automatic Stock Tracking

YOLO-based object detection runs at 15–30 FPS on an NVIDIA Jetson edge device, identifying and counting items continuously without any manual input from staff.

Manager Digest, Every Morning

A plain-English summary lands in your inbox before service begins — quantities on hand, items running low, and anything that expired overnight.

Anomaly Alerts

If a walk-in door is left open, an item disappears unexpectedly, or stock drops below a configured threshold, your team gets notified before it becomes a problem.

The Hardware

Industrial grade Edge Computing

The V1 Sensor is built to withstand the harshest kitchen environments while running complex neural networks in real-time.

Processor

NVIDIA Jetson Orin Nano

Sensing

ToF LiDAR + 4K RGB

Durability

IP67 Waterproof

Connectivity

WiFi 6E + LTE Failover

Thermal-resilient • Adaptive inference • Multi-sensor fusion

RL Training Lab

You are the reward signal.

Label real frames from our pokebowl dataset. Every correct answer sharpens the model. Every mistake teaches it boundaries.

10 frames. 10 decisions.

Each image was captured by our Jetson sensor inside a real pokebowl walk-in cooler. YOLO has already detected objects — now you decide what the model should learn.

Accuracy

Track your precision

Loss

Lower is better

Reward

Earn positive signal