EntroMetrix uses physics and AI to optimise industrial production, so existing plants deliver more output with less waste. Energy and material use fall, unplanned downtime drops, and throughput rises.
AI is a turning point for manufacturing, but only when it is grounded in engineering reality. Our mission is to establish a new ManufacturingOS that manages profitability, resource productivity and lower emissions as one system.
Industrial systems have outgrown the tools that run them.
A manufacturing site runs as one connected system of production assets, energy infrastructure, material flows and operating constraints. A small change in process stability, equipment performance or demand never stays local. It cascades across throughput, cost, efficiency and emissions, and the tools running the plant see each effect in isolation.
We build a working simulation of your process, calibrated against your live plant data, and run it alongside the site. The cascade becomes visible, so it can be optimised as a whole rather than one metric at a time.
A model trained on data alone is only reliable where it has seen data.
Plant conditions drift with new fuel mixes, demand patterns and equipment wear. Both models below learn from the same measurements. Where the data ends, one guesses. The other is held to the governing equations of kinetics, thermodynamics and energy balance, so it stays accurate in conditions it has never seen.
Optimisation coordinated automatically. Improved continuously.
Our science, packaged into a lean management system for the whole site. One continuous improvement loop your operators run every shift, taking waste out of every step of the process.
No rip-and-replace. We connect to the systems you already run, SCADA, IoT sensors, MES, historians and ERP, learn from your live process data, and send recommendations back through the screens your operators already use. First value in weeks, not a turnaround.
Throughput, energy cost per tonne, quality, waste and emissions are one coupled system, so pushing one lever moves the others. The model encodes your site's thermodynamics and operating constraints, so every recommendation carries the trade-offs your best operators hold in their heads.
Every setpoint stays inside the envelope your engineers define, with the reasoning attached. Operators accept, adjust or override, and the model learns either way. Nothing moves without a person, or a permission you have explicitly granted.
One operating model of the whole plant.
The model observes the plant end to end, finds the operating constraint, models the scenario, and rebalances flow, before anyone touches a setpoint.
Already deployed in production.
That loop is running live today, in three sectors.
We work with both large industrial organisations and smaller family-run manufacturers to reduce waste and energy intensity, improve operational efficiency, strengthen supply chain performance and lower operational emissions. These are the foundations of a lean operation.
AI and reinforcement learning, grounded in process physics.
Our models stay reliable outside the range of measured data because your process physics is built into them. Reinforcement learning with process physics, physics-informed Bayesian methods with uncertainty quantification and machine-learning-accelerated simulation are part of a wider stack we bring to each plant.
Read the science →Coverage and recognition.
Work on systems that directly impact efficiency, profitability and sustainability.
We are looking for individuals who are comfortable working across disciplines, from process engineering and control systems to machine learning and large-scale software architecture.
