Services
Control systems and automation engineering
Yunify is a product. This is the engineering work that sits either side of it: designing the control architecture a plant runs on, improving the one it already has, and connecting what the plant measures to what the business needs to know.
What the work covers
Four lines, and most engagements draw on more than one. They are separable because plants buy them separately, and a plant that needs its regulatory layer tuned does not necessarily want a digital twin in the same year.
Control architecture and strategy
PLC and DCS architecture for new plants, and architectural review for plants already running. Signal and interlock philosophy, alarm rationalisation, redundancy, segregation of the control and safety layers, and the interfaces a later analytics layer will need.
Advanced and AI-enabled control
Model predictive control, feedforward and constraint handling where a well-tuned PID loop has stopped being enough. Loop performance assessment first, because a plant with poorly tuned base regulatory control does not need an advanced layer on top of it.
Digital twin, IIoT and energy
Data acquisition from the interfaces a plant already exposes, historian and reporting architecture, and plant models that stay aligned with the asset. See what a digital twin for a power plant means in practice, and how to read plant data without modifying the control system.
Commissioning and troubleshooting
Loop checking, functional testing, and diagnosis of behaviour nobody can explain from the trends alone. Usually the first engagement, because it is where a plant discovers whether its instrumentation says what it thinks it says.
Why this sits alongside a product company
The two feed each other. Yunify is built on the assumption that plant data can be read, trusted and interpreted in operating context, and every one of those assumptions is tested by control engineering work on real plants rather than by a roadmap.
It also means an engagement does not have to start with a product decision. A plant that wants its specific energy consumption explained does not need to buy an operating intelligence layer to get an answer. Often the useful first step is establishing whether the measurement boundary and the instrumentation can support the question being asked at all.
It also runs in the other direction. On a troubleshooting engagement the Engine can be brought in as an instrument rather than as a purchase. That can be a thermodynamic model pointed at the plant's own historian to explain where the heat or the energy is going, the condition-monitoring models for rotating machines pointed at a machine nobody can account for from the trends alone, or the anomaly models run against live data to find a logic or interlock problem that only appears in a particular plant state. The plant gets the answer either way; whether it then wants that layer running permanently is a separate decision, taken later.
Where a plant does go on to deploy Yunify, the same people who mapped the control architecture are the ones who know which tags mean what, which is the part that usually takes longest.
How an engagement runs
Scope first, in writing, against what the plant can actually expose. That means agreeing the interface, the data, the access route and the change-control position before anything is connected, and confirming the proposed connection against the OEM warranty position and the site's management-of-change process.
Nothing writes to control logic. Reading is done through an existing northbound interface wherever one is available, and any field-level integration is treated as a separate engineering change with its own review rather than folded into a monitoring scope.
The safety layer stays out of it. IEC 61511 requires the safety instrumented system to be independent of the basic process control system to the extent that its integrity is not compromised, and no analytics or advisory function is placed where a safety function would come to depend on it.
Who does the work
Control engineering with certification behind it rather than a general software team learning the domain. The practice is led from inside the company: ISA Certified Automation Professional and TÜV-certified functional safety experience, with a background in advanced control and multi-physics modelling at GE, Mitsubishi and Alstom.
About Yuji Labs covers the company, where it operates and the credentials of the two founders in full.
Frequently asked questions
Do you work on plants that will never deploy Yunify?
Yes. Control architecture, loop performance and commissioning work stands on its own and is scoped that way. There is no requirement to adopt the platform, and the engineering findings belong to the plant either way.
Which control systems do you work with?
PLC and DCS platforms in common industrial use, reached through the interfaces the plant already exposes: OPC UA or DA, Modbus TCP, MQTT, historian interfaces and OEM APIs. Where a required interface is not available, that becomes a scoped engineering change rather than an assumption.
Can advanced control be added to an existing plant?
Often, but the assessment comes first. Loop performance, valve condition, measurement quality and constraint definition all determine whether a model predictive layer will hold. A plant whose base regulatory control is poorly tuned usually gets more from fixing that than from adding a layer above it.
Does connecting an analytics layer affect the equipment warranty?
That is a contractual question rather than a technical one, and it is answered by the OEM agreement and the site's management-of-change procedure, not by a generic architecture. The engineering position is documented so the plant can put it to the vendor and get a written answer before anything is connected.
Which sectors does this cover?
Green hydrogen and electrolyser plants, battery storage, solar PV and concentrated solar, wind, thermal and conventional gas power, desalination, and downstream compression, storage and ammonia. The control problems are more alike across those than the equipment suggests.