Quality inspection is still sampling
You check one unit in fifty and hope the other forty-nine were fine.
Home / Industries / Manufacturing
Manufacturing
Inference on the line, not in someone else's datacentre.
Vision inspection, predictive maintenance and plant analytics that run at the edge — so a network outage slows the internet, not the line.
The three things we hear in the first meeting.
If none of these sound like you, we are probably not the right call yet — and we will say so.
Downtime is discovered, not predicted
The maintenance schedule is a calendar, not a signal from the machine.
The plant network cannot be trusted with production
Anything that needs a round trip to the cloud will stop the line the day the link drops.
If it has to stop the line, it has to run on the line.
Plant-floor AI has a constraint most software does not: the fallback cannot be 'wait for the network'. We design for local inference first and treat the cloud as where the reporting goes.
-
Runs at the edge.
Inference on local hardware next to the line, with the cloud used for aggregation and dashboards.
-
Degrades safely.
If the uplink drops, inspection continues and syncs when it returns.
-
Built around OT reality.
We work with your existing PLCs, cameras and SCADA rather than asking you to replace them.
-
Measured against the current baseline.
We record what your defect and downtime numbers are today, so the improvement is not a claim.
Who we build for
Different teams, different first project. The platform underneath is the same.
Quality teams
Vision inspection on every unit, with defect classes your team defines and can retrain.
Maintenance and reliability
Vibration, thermal and current signals turned into a maintenance signal rather than a calendar.
Plant management
One view of OEE, scrap and downtime across lines and sites.
Plant IT and OT
Edge hardware, networking and the managed layer that keeps it patched and monitored.
What we actually deliver
Four things, in this order. Each one is useful on its own.
Line-side discovery
A week on the floor. We come back with the inspection or downtime problem worth solving first.
Vision inspection system
Cameras, lighting, model and the operator interface, trained on your defects.
Edge inference stack
Local compute sized to the line, with safe degradation and automatic sync.
Plant analytics
Scrap, throughput and downtime reporting that plant and group management both trust.
How a manufacturing engagement runs
No surprises about sequence, and no invoice before there is something to look at.
-
01
Floor walk and baseline
We measure current defect escape and downtime before proposing anything.
-
02
Proof on one line
One line, one defect class, real product. Measured against the baseline.
-
03
Harden for production
Edge hardware, failure modes, operator training and the retraining loop.
-
04
Replicate across lines
The second line takes a fraction of the time. The tenth is a rollout, not a project.
What good looks like
- 100%
- Units inspected, not sampled
- Edge
- Inference local to the line
- Offline-safe
- Keeps running through network loss
- Retrainable
- Your team adds new defect classes
The Manufacturing AI readiness checklist
The questions we work through before recommending anything — data readiness, hosting constraints, review process and the running cost at year two. Use it with any vendor.
Manufacturing questions we get asked first
Short answers. Longer ones are a conversation.
Do we need to replace our cameras?
Often not. We assess what is installed during the floor walk. Sometimes lighting is the real problem and the camera is fine. Where a sensor genuinely cannot support the inspection, we say so and specify what would.
What happens when the network goes down?
Inference runs on local hardware, so inspection continues. Results queue locally and sync to the dashboard when connectivity returns. Nothing on the line waits for the cloud.
How many defect samples do you need to train?
Fewer than most people expect for obvious defects, more than most expect for subtle ones. We give you a realistic sample requirement after seeing the defect classes, and we can start with synthetic augmentation where real defect images are scarce.
Can our team retrain it, or do we call you every time?
Your team retrains. Part of the handover is the labelling and retraining workflow, because a system only your vendor can update is a system that goes stale.
Will this integrate with our MES or SCADA?
Yes, usually via OPC UA, MQTT or a direct database write, depending on the vendor. Integration is scoped during discovery.
Tell us which line is costing you the most.
Scrap rate, downtime hours or inspection labour — start with whichever number hurts.
- A reply within one working day, from an engineer rather than an account manager.
- An honest read on whether this is worth doing now, or in a year.
- No marketing list. Your details reach the solutions team and stop there.