DRAGONFLY · VISUAL QUALITY CONTROL

Catch a bad part
before it leaves the machine.

High-resolution visual anomaly detection for the production line. Give the reference part and its measures, point a camera at what comes off the line — let the model tell you if something is wrong.

Start a Trial ↓
Open the Dashboard →

No training required — a pretrained vision LLM reads the part against its spec.

How it works

Quality control in three steps

No model to train, no data to label. Define the reference, capture the part, get the verdict.

01

Define the reference

Upload the ideal part image and its reference measures once. The spec is locked in from there.

02

Capture at the station

A camera on each machine streams a high-resolution view of the produced part from every angle — automatically, as it comes off the line.

03

Get the verdict

The vision LLM compares the part against the spec and tells you where a deviation is, if any — and raises an alarm.

Put your line to the test

Upload your first part and see the model at work. No setup, no training — give the reference, get the deviation.

Early access opens in turn for teams running a production line.