CHALLENGE · 01
OT/IT convergence · without breaking either
OT engineers don’t trust IT. IT doesn’t understand OT. We sit in the middle with a Purdue-model boundary that respects both.
ISA-95 · IEC 62443Industry · 04 of 06 · Discrete · Process · Industry 4.0
The factory sees itself. The line tells you before it stops.
A plant doesn’t care about your software stack. It cares about uptime, yield and the next shift. We sit on the OT/IT boundary — wiring sensors, MES, ERP and vision systems into a single live picture of the floor, then layering ML that earns its keep in fewer breakdowns.
−41%Unplanned downtime · YoY
99.7%First-pass yield · vision QC
+6.2 ppOEE lift · 12 months
plant · A14 · line group · 3 of 3OPC-UA · LIVE
S1
S2
S3
S4
S5
S6
Bearing S4 · vibration RMS up 38% / 6h · predicted failure in 18–26h · WO drafted
99.7%First-pass yield
OEE 86.4Plant · 7d
−41%Downtime · YoY
The reality on the ground
What an operating plant actually wants from software.
Not a digital-twin slide. Not a dashboard. Not a six-month BI engagement. A working answer to “why is line 3 down?” — before the shift supervisor has to walk over and ask.
FACT · 01$50B
Lost annually to unplanned downtime in industrial manufacturing. Most of it is preventable with signals the plant already produces.
FACT · 0280%
Of factory data never leaves the PLC. Stranded on OT networks that don’t talk to IT — and were never meant to.
FACT · 03OEE 64%
World-class is 85%. The gap is rarely a machine problem. It’s a visibility problem.
FACT · 0415 yrs
Average age of MES installs still running production. We don’t replace them — we connect them.
The hard problems
What keeps a Plant Director & Head of Digital awake at 02:00.
Every plant has the same handful of issues, named slightly differently. Here’s the version we hear from operations on day one.
CHALLENGE · 02
Reactive maintenance is too expensive
A failed bearing is a $4,000 part. The shift it takes down is a $400,000 day. Predictive is no longer optional.
Vibration + acoustic MLCHALLENGE · 03
Quality at line speed
Manual QC samples 5% of output. Computer vision can inspect 100% — without slowing the line.
Edge inferenceCHALLENGE · 04
MES · ERP · supply-chain mismatch
SAP says one thing, the floor says another, the shipping label says a third. Finance reconciles by hand.
Event-driven syncCHALLENGE · 05
Energy & sustainability reporting
CBAM, SECR, CSRD. Scope 1/2/3 data with audit-grade lineage — directly from the meters, not a spreadsheet.
CSRD · Scope 1/2/3CHALLENGE · 06
Workforce knowledge walks out the door
The shift supervisor who knew the line retires. The runbook lived in their head. Now it doesn’t.
AR + tribal-knowledge AIWhat we build · and own
Six modules. One industrial backbone.
We meet the plant where it is — Modbus, OPC-UA, EtherNet/IP, PROFINET, MQTT — and bring the cloud to the shop floor rather than vice versa.
Module · 01
IIoT & digital twins
OPC-UA / MQTT ingestion from PLCs, drives, robots and edge gateways. A live, queryable twin of the line that runs alongside the physical one.
- OPC-UA · MQTT · Modbus
- Edge gateway fleet mgmt
- Live equipment state
- Twin API for engineers
Module · 02
Predictive maintenance
Vibration, current, acoustic and thermal signatures fed into failure-mode models. Bearings, motors, gearboxes — caught 18–72 hours ahead.
- Vibration RMS + crest
- Current signature analysis
- Spectral anomaly ML
- Auto-drafted work order
Module · 03
Computer-vision quality
Camera-based defect detection at line speed. Pull bad parts before they leave the cell — and feed the upstream process the signal to self-correct.
- 100% inspection rate
- <50ms inference at edge
- Closed-loop feedback
- Defect taxonomy training UI
Module · 04
MES · ERP · supply-chain glue
SAP, Oracle, Plex and homegrown MES all speak different dialects. We translate, reconcile and expose a clean event stream that finance, planning and ops can all trust.
- SAP iDoc · BAPI
- Plex / Rockwell connectors
- Idempotent ETL
- Reconciliation UI
Module · 05
OEE & shift analytics
Real-time OEE per asset, shift, product, operator — drilled down to the six big losses. No more weekly reports that arrive a week late.
- Six-losses decomposition
- Per-shift comparison
- Andon escalation rules
- Mobile shift dashboard
Module · 06
Energy · sustainability · CSRD
Meter-grade Scope 1/2/3 with audit lineage. CBAM reporting, embedded-carbon per SKU, real-time efficiency scorecards.
- ISO 50001 ready
- CSRD · SECR · CBAM
- Per-SKU embedded carbon
- Anomaly alerting
Reference architecture
From PLC tag to business action, the path is six steps.
The same event — a sensor reading, a quality reject, a planned downtime — flows through six clean stops on its way to becoming a shift supervisor’s next move.
STEP · 01SensePLCs, drives, vision, energy meters exposed via OPC-UA / MQTT.
STEP · 02EdgeEdge gateways aggregate, filter, run inference under 50ms.
STEP · 03StreamTime-series + event store in TimescaleDB · Iceberg cold.
STEP · 04ReasonModels predict failures, score quality, calculate OEE.
STEP · 05SurfaceAndon, shift dashboard, work-order draft, supervisor mobile.
STEP · 06SyncERP, MES, CMMS, planning notified · audit trail sealed.
Tools we reach for
Battle-tested stack. No buzzword tax.
OT · industrial protocols
OPC-UAMQTT 5Modbus TCPEtherNet/IPPROFINETSparkplug B
Edge & platform
Azure IoT EdgeAWS GreengrassK3s · FleetTimescaleDBApache IcebergKafka
AI / ML
PyTorch · TensorRTTriton inferenceAnomalibONNX RuntimeMLflowEdge Impulse
Application & ERP
SAP S/4HANAPlexRockwell PlantPAxIgnition · InductiveReact Native
Plant types we’ve worked in.
Discrete and process, continuous and batch — from auto-tier suppliers and electronics fabs to food, pharma and specialty chemicals.
Automotive · tier 1/2Electronics & semisFood & beverageSpecialty chemicalsPharma manufacturingHeavy machineryAerospace & defenceConsumer packaged goods
Case study · in production
How an auto-tier-1 supplier paid for the whole programme in one quarter.
Three plants, eleven thousand SKUs, and an OEE figure nobody believed. We instrumented the lines, layered vibration-based predictive maintenance and brought the MES feed into the same dashboard. Bearing-fault prediction saved one weekend in quarter one — and the financial case was closed.
ClientAuto tier-1 · 3 plants
Assets412 critical machines
Timeline10 weeks · pilot to first plant
Team6 ETY · 5 client OT engineers
−41%Unplanned downtime · YoY
+6.2 ppOEE lift · 12 mo
99.7%First-pass yield · vision QC
1 wkdSaved · paid the programme
Compliance · built-in
Audit-ready by default. Not by sprint.
Industrial cyber is the new audit frontier. We build to IEC 62443 segmentation from day one, never bridge OT to IT without a documented boundary.
IEC 62443OT cyber · zones + conduits
ISA-95L0–L4 functional hierarchy
ISO 27001 / 27017ISMS · cloud-extended
GxP · 21 CFR Part 11Pharma manufacturing
CSRD · SECR · CBAMSustainability disclosures
NIST CSF 2.0Critical-infra baseline
ISO 50001Energy mgmt systems
CMMC L2 · ITARAerospace · defence
Questions you’ll probably ask
The short version of the kick-off call.
Do we have to replace our existing MES / SCADA / historian?+
How do you handle OT cyber without breaking the firewall?+
How accurate is predictive maintenance, really?+
Can computer-vision QC actually keep up at line speed?+
What about brownfield equipment with no native sensors?+
How do you handle our CSRD / Scope-3 reporting?+
Plant · OT · Industry 4.0
Bring us to the plant floor. We’ll bring the runbook.
A 60-minute working session with a senior engineer and an OT lead. We’ll walk one of your lines (virtually, if needed), pull the data you already have, and show you the three quickest wins.