Machine Pilot AI
Machine Pilot Vision™ · In beta · Coming soon

Machine health and performance, visible like never before.

Next-generation technology. Machine Pilot Vision™ is a standalone industrial-camera AI that watches your machinery continuously, around the clock. It captures and logs machine activity to detect both gradual performance degradation and sudden faults — delivering real-time alerts and the machine-health and performance metrics that simply weren't visible before.

What Machine Pilot Vision delivers

Real-time machine health — captured, tracked, and connected.

Machine health in real time

Continuous visibility into how every monitored machine is performing right now — not after the shift, not from a spot check.

Gradual and sudden change, logged

Slow performance drift and abrupt faults alike — captured and time-stamped so every change is tracked over time.

Insight that wasn't measurable before

Surfaces performance data and patterns no manual check or fixed sensor could capture.

One platform, all your intelligence

Vision data feeds straight into the Machine Pilot platform — live visual telemetry alongside your documents, PLC data and AI, in a single view.

Behavioural deviation detection

Built around change from normal — not a fixed list of faults.

Rather than hard-coding "detect belt slip" or "detect chain stretch", Vision learns what normal looks like for your machine and watches for movement away from it.

Timing changes Motion changes Sequence changes Throughput changes Position changes Flow changes Robot path changes

What it tells you

  • What changed
  • When it changed
  • How severe the change is
  • Whether it's progressive or sudden

How experienced people reason

This mirrors how a senior maintainer actually thinks about a machine — not "which fault code is this?" but "that doesn't sound, move or time the way it should." Vision turns that instinct into a measurable, logged signal.

Progressive vs sudden

Every deviation is classified by how it arrived.

Knowing whether a change crept in or appeared instantly is what makes the signal useful for maintenance planning.

Progressive deviation

A gradual drift over time — the early signature of wear.

  • Component wear
  • Belt deterioration
  • Chain stretch
  • Bearing degradation
  • Product-flow deterioration

Sudden deviation

A rapid step away from established operation.

  • Incorrect setup
  • Sensor failure
  • Product jam
  • Mechanical failure
  • Robot collision
  • Parameter changes
Visual observation network

Cameras placed where the data is.

Depending on machine complexity, monitoring uses a single camera or several — with dedicated views of the critical mechanisms, product flow and robotic systems where performance is won or lost. Each camera is positioned to capture the key data a machine's condition depends on.

Product infeed Product discharge Belt drives Chain drives Pusher systems Timing mechanisms Conveyor transfers Robotic pick points Palletising areas

The objective: capture the visual data a machine's condition and performance depend on — continuously, across every critical zone.

How it learns

Learn the baseline, then compare against it — continuously.

Phase 1 · Baseline learning (~2–3 weeks)

After install, Vision watches the machine run and builds a behavioural model of normal — adjustable to machine complexity.

  • Records machine operation
  • Learns movement patterns & sequences
  • Learns cycle timing & throughput
  • Establishes normal operating ranges
Phase 2 · Continuous comparison

Once a baseline exists, all future operation is measured against the learned model, second by second.

  • Timing & motion variation
  • Positional variation
  • Process variation
  • Throughput variation

When operation moves outside accepted ranges, the event is classified, scored for severity and logged.

Process-flow intelligence

OEE-style insight — visually, without touching your PLC.

Vision can identify and track products, cartons, containers and pallets right through a production sequence — measuring performance independent of your control system.

Incoming carton
Product insertion
Carton closure
Carton transfer
Palletisation

Measured without PLC access

  • Cycle times
  • Queue lengths
  • Waiting times
  • Throughput
  • Flow interruptions
  • Process bottlenecks

Delivering OEE-style insight independent of the customer's control system is a significant commercial differentiator — no integration project required.

Event response framework

When something changes, the evidence is already captured.

Log the event
Capture video evidence
Save pre-event footage
Save post-event footage
Highlight the affected area with overlays
Record deviation severity
Notify designated personnel
Store the event in the Machine Pilot platform
Dashboard Email SMS
Architecture · Version 1

Edge-first, on its own network — independent of plant IT.

Cameras feed a local NVR and an edge processing server on a dedicated Machine Pilot network. The edge device does the heavy lifting; only metadata and the clips that matter reach the cloud.

Industrial cameras
Dedicated Machine Pilot network
Local NVR / recorder
Edge processing device
Secure internet connection
Machine Pilot Cloud → dashboard & alerts

On site — the edge server

Video decoding, motion tracking, object detection, timing analysis and event detection all happen locally.

In the cloud

Event logs, performance metrics, alert data, short clips and behavioural models — powering long-term analytics, cross-machine comparisons, dashboards, alert management, historical reporting and AI model management.

Why edge-first matters

Streaming raw video — say 8 cameras × 1080p × 24/7 — to the cloud creates enormous bandwidth and storage costs. Processing at the edge keeps most video on site and sends only what's important.

  • Independent of plant IT
  • Faster, simpler rollout
  • Reduced cybersecurity concerns
  • Lower latency & better reliability
  • Consistent across customers
  • Easier to scale
Performance

Designed for near real-time monitoring.

Low video latency, rapid event detection and reliable continuous operation. Specific latency targets are deliberately left to be established empirically during technical design rather than quoted before they're proven.

In beta · Coming soon

Want to be an early Vision site?

Machine Pilot Vision™ is in beta. If you've got a critical machine you'd watch by eye if you could be there every shift, we'd like to talk about putting it on camera.

Register your interest

This page describes a product roadmap currently in beta. Capabilities and timelines may change as the platform develops.