Your best inspector can't be at every station. What he knows can be.

Weld and in-line quality inspection that runs at the cell, catches the escape while it is still cheap to fix, and keeps your data inside your building.

What it costs you

Three costs, in the order manufacturing teams actually raise them.

Escapes surface late

A defect found downstream costs hours of rework as the unit moves through the line, not the minutes it would have cost at the cell. The further it travels, the more of the build has to come apart.

Where Nyx helpsDetection immediately after the process step, while remediation is still a rework of one operation rather than of an assembly.

The standard walks out the door

What good looks like lives in a handful of experienced people. It is not written down in a form a machine can apply, and no one of them can be at two stations at once.

Where Nyx helpsThe standard captured once from your own people and your own parts, then applied at every station, with the metrics that show it holding.

Manual inspection does not scale

Fixtures and flashlights confirm that a weld is present. Presence is not quality, and the gap between the two is where the escapes live.

Where Nyx helpsConsistent classification at line rate, on your hardware or ours, running locally with no cloud dependency and no data leaving the plant.

What this looks like on the line

One operation, instrumented. The loop matters more than any single step.

WeldYour existing cell and process. We do not touch it.
CaptureCameras and lighting positioned for the defect modes you care about.
DetectClassification at the cell, locally, in the time the operation already takes.
DispositionPass, rework, or hold, with the evidence attached to the unit.
ActRework at the station, and the record kept for the quality documentation you already owe.

Dispositions feed back into the model. Where the machine and your inspectors disagree, that disagreement is the training signal, and detection accuracy improves against a baseline agreed with you before anything was deployed.

The trade-off we will put in front of you

Your hardware is cheaper. Ours is better instrumented. You should decide, not us.

Running on cameras and compute you already own is the lowest-cost path and it keeps the capital argument simple. It also means the sensor quality and the inference budget are fixed before we start, and some defect modes may sit below what that stack can resolve.

A Nyx-defined unit lets us control sensor quality and inference together, which is what it takes to reach the harder defect modes. It costs more and it is a new line item.

We will tell you which defect modes fall on which side of that line, in writing, before you commit to either.

The first step

A bench-top prototype against your own parts, or production-intent units that you own outright. Both are scoped in a two-week evaluation that ends in a written assessment, a measured baseline, and a firm price.

Book an evaluation See what we have built