A standard Industry 4.0
architecture, end-to-end.
Virtual DX showcases CGI's Reference Architecture for industrial digital transformation — a vendor-agnostic, scalable blueprint that connects shop-floor reality to enterprise systems and AI, the way a production-grade i4.0 platform should.
The i4.0 problem we solve
Eight steps. One platform.
Every industrial transformation needs to do the same eight things, in the same order. Most organizations stitch them together with point-to-point integrations and end up locked in. The CGI Reference Architecture replaces that chain with a single, decoupled platform.
The Reference Architecture
Four platform components, plus a delivery layer on top
The CGI Reference Architecture splits the platform into four loosely-coupled, swappable components — IIoT, DataOps, Broker, Historian — that together expose an actionable digital twin of the business. On top sits an AI & application-development layer that lets anyone in the business build their own solution to their own problem or opportunity, without touching the platform itself.
IIoT — Connect
Reach the asset wherever it is. Speak the protocols of the floor (OPC-UA, Modbus, S7, MQTT, Sparkplug B), respect OT segmentation, and stream raw signals upward at the right rate and quality.
raw.py (paho-mqtt) · 169+ synthetic tagsDataOps — Contextualize
Turn raw signals into actionable information. Transform, enrich, normalize against an ISA-95 model, reconcile with ERP / MES events, and make every value mean the same thing across sites and applications.
HighByte Intelligence Hub 4.4Broker — Expose the UNS
A single Unified Namespace acting as a live digital twin of the business. Producers publish once; any consumer subscribes — no point-to-point integrations, no vendor lock-in.
HiveMQ CE · MQTT 5 · WS + TCP · retainedHistorian — Persist
A time-series record of every UNS value, indexed by ISA-95 tags for slicing across sites, lines, shifts, and products. The foundation for trends, BI, ML training, and audit.
InfluxDB 3 Core · SQL queriesAI & application-development framework
The platform exposes a stable, semantic surface. On top of it, an AI & low-code application-development layer lets anyone in the business — process engineers, maintenance leads, line managers — build their own solution to their own Problem & Opportunity. Dashboards, copilots, predictive models, automations: every one of them consumes from the same UNS, never touches the OT layer, and ships independently.
React dashboard · OpenAI copilot · SAP S/4 simulator · ready for low-code citizen apps
Why this matters
Business outcomes the architecture enables
The four-layer split isn't an academic exercise — it directly produces the outcomes our clients ask for. Some take effect day one; others compound over the lifetime of the platform.
Decouple data from applications
No more point-to-point integrations. A new app subscribes; nothing else on the floor knows or cares.
Democratize data access
One namespace, one model, one source of truth. Operators, analysts, data scientists and partners read the same numbers.
Enable scalable AI
Clean, contextualized, time-stamped data — the precondition for ML and LLM-based copilots that actually work in production.
Real-time visibility
Sub-second OEE, alarms, downtime — the same data flow used for dashboards is reused by automated workflows.
Vendor-agnostic, future-proof
Open protocols (MQTT, OPC-UA, REST). Swap any single component without touching the rest. No lock-in.
Lower integration cost
N apps × M sources collapses from N×M custom connectors to N+M. Each new endpoint is incremental, not exponential.
Security & governance by design
One gateway, one identity perimeter. OT stays on its own segment — applications never touch the control network.
Operational autonomy
Site teams keep ownership of their data. Central teams build platforms, not pipes.
How a signal travels
From a casting furnace to an LLM, through the four components
Pick any tag in the live dashboard. This is the path it took.
raw/Group_A/0x19BC02.uns/ACME/Lyon/Casting_Center/Billet_Line_01/Furnace_01/....↑ On top: dashboards, the copilot, and any other app subscribe to the broker (live state) and query the historian (history) — none of them knows or cares what's behind it.
Tech stack
Reference-architecture components used here
Every choice below is interchangeable — the architecture survives the swap.
Ready to see it?
Walk the full path: a synthetic alarm fires, the dashboard updates, the copilot reasons over it, SAP gets a notification — all in real time.
Open the demo →