Software to see what is happening in a plant, understand why and decide in time.
The data does not leave the plant. Each plant has its own model, trained on its own data, and it does not depend on the cloud to run.
Animation: the camera pulls back from a sensor on a pipe. The process appears with its related signals; then the whole plant in dashed lines and, finally, several facilities of the water cycle.
Where we start
Animation: a water network is drawn from the tank, water runs through it and the level, pressure and flow sensors relate to each other.
Possible expansion
Animation of a possible expansion sector: a batch tank fills and is stirred; next to it, the expected temperature range changes with each phase and the measured signal follows it.
Possible expansion
Animation of a possible expansion sector: a slab leaves the furnace and thins as it passes through two rolling stands; the temperature, vibration and energy sensors relate to each other.
60% of Spain's population is already within the scope of the plan that digitizes the water cycle (PERTE).
Source: iAgua, 13 May 2026
NIS2 The European directive classes drinking water and waste water as sectors of high criticality.
ES-ISAC Aguas The water sector already shares cybersecurity information with INCIBE.
Source: INCIBE, 11 Jun 2026
02What we do
Five steps, from the first reading to the first diagnosis.
02What we do
Five steps, from the first reading to the first diagnosis.
The state of the plant: sensor measurements and actuator states. It only reads: it never touches the control system.
Signals do not change on their own. When the pump starts, the outflow rises and the tank level drops. A signal makes sense next to the others.
It learns how the signals move together from normal operating data of the plant itself, with no examples of faults. At every moment it checks whether what it sees fits what is expected.
The pump and the flow keep their rhythm, but the level no longer drops as it should. Everything is still within limits and looks normal at a glance. ATLAXIA flags it as an anomaly; a fixed threshold would wait for the level to cross it.
A first diagnosis when something moves away from normal: which signal and which process, with an anomaly level from 0 to 5 per sensor, per process and for the plant. It also alerts when a signal crosses the operator's limits. The decision is still yours.
Under the hoodA graph neural network treats each signal as a node and learns which signals influence each other. A model you can understand matters as much as one that gets it right: every alert is explained in terms of your plant's equipment, processes and signals.
The state of the plant: sensor measurements and actuator states. It only reads: it never touches the control system.
Signals do not change on their own. When the pump starts, the outflow rises and the tank level drops. A signal makes sense next to the others.
It learns how the signals move together from normal operating data of the plant itself, with no examples of faults. At every moment it checks whether what it sees fits what is expected.
The pump and the flow keep their rhythm, but the level no longer drops as it should. Everything is still within limits and looks normal at a glance. ATLAXIA flags it as an anomaly; a fixed threshold would wait for the level to cross it.
A first diagnosis when something moves away from normal: which signal and which process, with an anomaly level from 0 to 5 per sensor, per process and for the plant. It also alerts when a signal crosses the operator's limits. The decision is still yours.
Under the hoodA graph neural network treats each signal as a node and learns which signals influence each other. A model you can understand matters as much as one that gets it right: every alert is explained in terms of your plant's equipment, processes and signals.
03What we are building
Technically, a multi-layer ontology: what is in the plant and how each part relates to the rest, written so that the model and people understand it the same way.
Three layers describe the plant; the model learns the fourth.
03What we are building
Technically, a multi-layer ontology: what is in the plant and how each part relates to the rest, written so that the model and people understand it the same way. Three layers describe the plant; the model learns the fourth.
What is in the plant: processes, equipment and tanks, and the signal from each sensor.
Where the data travels: controllers, SCADA and communications. The network can be listened to without interfering with it.
How the process is chained: which stage feeds which, and the rules engineering knows.
The one the model learns on top of the other three: which signal influences which, how strongly and with how much delay. A learned relation is a lead, not a proven cause.
Sketch of a pumping station: tank, pump and the line to the network. Four sheets are laid over it. Physical: intake, pumping and distribution, with six instruments linked to the point they measure. OT network: PLCs, an RTU, the SCADA, the historian and ATLAXIA, which receives its data. Functional: the path of the water, from the inlet to the tank, the pump, the flowmeter and the valve, out to the network. Influences: the plan tilts back and, above it, a second sheet with one node per instrument and the relations between them, thicker the stronger they are.
What is in the plant: processes, equipment and tanks, and the signal from each sensor.
Where the data travels: controllers, SCADA and communications. The network can be listened to without interfering with it.
How the process is chained: which stage feeds which, and the rules engineering knows.
The one the model learns on top of the other three: which signal influences which, how strongly and with how much delay. A learned relation is a lead, not a proven cause.
Sketch of a pumping station: tank, pump and the line to the network. Four sheets are laid over it. Physical: intake, pumping and distribution, with six instruments linked to the point they measure. OT network: PLCs, an RTU, the SCADA, the historian and ATLAXIA, which receives its data. Functional: the path of the water, from the inlet to the tank, the pump, the flowmeter and the valve, out to the network. Influences: the plan tilts back and, above it, a second sheet with one node per instrument and the relations between them, thicker the stronger they are.
04Who we are
A team formed in UCAM research projects and a technology already tested in real water facilities.
04Who we are
A team formed in UCAM research projects and a technology already tested in real water facilities.
We started at the university. The ATLAXIA team formed at UCAM, collaborating on a range of research projects. Those projects are the technical foundation of the company.
A UCAM spin-off. ATLAXIA is a project built on a university spin-off. Research is the starting point; the plant is where it has to work.
The technology comes from TRUEDATA. Its name stands for secure and reliable traceability of water infrastructure data. It is a project in INCIBE's pre-commercial public procurement program, the way public administrations commission the development of solutions that do not yet exist on the market.
It was tested in three real facilities. A drinking water treatment plant, a wastewater treatment plant and an industrial plant with process water, with their own data and operating conditions. One of them had no Internet or remote access: everything had to run on site, without touching what was installed.
From protecting data to understanding the plant. TRUEDATA was born to protect the data, with the cybersecurity of the control network at its core. While building the platform we saw a wider reach: the model of equipment, connections and processes could explain the whole plant, not just its network.
ATLAXIA is its commercialization stage. What worked in TRUEDATA becomes a product: it runs on a unit at each plant, learns from the plant's own data and works alongside the systems already there. The data never leaves the plant.
We build it with the people who run the plants. Every integration is tuned with the plant's operators. We are looking for water plants to build it with, starting with a scoped trial.
Team
Research, artificial intelligence and product. Three profiles that cover the whole path: from the idea to the model, and from the model to the plant.
CEO
CTO and AI development
CMO and full-stack development
Whether you run a plant, do research, work with industry or want to see the product running.
Request a demoOr write to us about
Emailhola@atlaxia.es