Plant software that improves the line without stopping it
We build edge-first industrial systems that capture why a machine stopped, link quality back to process, and deploy in stages that production can absorb.
Edge-first
capture at the source, survive the network
Reason-coded
downtime explained, not just counted
Zero-stop
rollout staged around production
Why plant data rarely reaches the people who need it
Industrial environments mix twenty-year-old equipment, half a dozen vendor protocols, safety constraints, and a line that cannot pause for an upgrade. Data exists but sits trapped in controllers and vendor historians, so downtime reasons are reconstructed from memory at the end of a shift. We design edge-first integration and staged rollout, improving decisions without ever putting output at risk.
- 01
Every machine speaks differently
PLCs, serial devices, and vendor historians each expose their own protocol and vocabulary. Normalising them is the unglamorous foundation of everything else.
- 02
Downtime is recorded from memory
When reasons are entered at the end of a shift, the data reflects recollection. Improvement work then targets the wrong causes with great confidence.
- 03
The line cannot stop for software
Change windows are short and safety cases are strict. Anything requiring a full production halt will not get approved, however good it is.
The build sheet for a connected plant
Every line item below runs in production somewhere. We can take one of them or the whole sheet, from signal survey through to fleet rollout.
- Spec 01
OEE and production visibility
Availability, performance, and quality computed from machine signals with reason codes captured where and when the stop happens.
- OEE
- Reason codes
- Shift reporting
- Spec 02
Edge and controls integration
OPC UA, MQTT Sparkplug, Modbus, and vendor drivers unified into one model, with store-and-forward for network loss.
- OPC UA
- MQTT Sparkplug
- Store and forward
- Spec 03
Quality and traceability
Inspection workflows and genealogy linking finished output back through process parameters to raw material lots.
- Genealogy
- SPC
- Inspection
- Spec 04
Maintenance and condition monitoring
Condition signals, thresholds, and work-order integration so maintenance is planned from evidence rather than a calendar.
- Condition monitoring
- Work orders
- Thresholds
- Spec 05
MES and ERP connectivity
Work orders, routings, and confirmations flowing between the floor and business systems without a nightly batch in between.
- MES
- ERP sync
- Work orders
- Spec 06
Industrial analytics and models
Anomaly detection and yield models trained on plant data, deployed at the edge where the decision has to be made.
- Anomaly detection
- Yield
- Edge inference
What we build plant systems with
Chosen for fidelity at the edge, survival through network loss, and one platform that can be operated across every site.
Plant surface
Readable from the floor
- React
- Next.js
- TypeScript
- Grafana
- Andon displays
Edge layer
Runs when the network does not
- Rust
- Go
- MQTT Sparkplug
- OPC UA
- Docker
Time series and data
High-frequency signals at scale
- TimescaleDB
- InfluxDB
- Kafka
- Apache Spark
- Snowflake
Platform
Fleet management for plants
- Kubernetes
- K3s
- Terraform
- Azure IoT
- Prometheus
From signal survey to plant-wide rollout
Commissioning order matters more than feature order. Each stage has to be provable on the floor before the next one is worth starting.
- 01
Survey the signals
We audit what each asset already emits, what is reachable, and what needs new instrumentation before promising any metric.
- 02
Prove it on one cell
A single line or cell becomes the reference implementation, including the reason-code taxonomy operators will actually use.
- 03
Harden for the plant floor
Store-and-forward, offline operation, and safe redeploy are validated against real network and power conditions.
- 04
Scale plant by plant
Edge deployment becomes a repeatable fleet operation, with configuration per site and one shared platform behind it.
What plant and operations teams ask us first
Usually yes. Older assets are reached through PLC tags, serial gateways, or added sensing where nothing is exposed. The survey stage establishes exactly what is possible per asset before anyone commits to a metric.
Make the line legible before you change it
Tell us which cell, line, or plant is costing you output. We will come back with a signal survey, a reference build, and a rollout your production schedule can absorb.