We built the tools we needed when we ran these plants.
Twynvex was founded in 2025 by a team of process engineers and simulation specialists who spent years frustrated by the gap between what physics-based process models could do and what was actually deployed on plant floors. We decided to close that gap.
"The enthalpy balance of your distillation column doesn't become more forgiving because nobody modeled it. Every continuous-process facility should have a running physics model of itself — one that updates every 15 seconds and tells you where the process is going, not where it has been."
Modern DCS systems generate enormous amounts of sensor data. But raw sensor data alone cannot tell you whether your reactor train is trending toward a yield deviation — only a model of the process physics can do that. Statistical models and ML-based approaches have been tried across the industry. They fail precisely when you need them most: when operating conditions shift outside the historical training distribution, which is exactly when a deviation is happening.
Twynvex uses first-principles models — the heat and mass balance equations that a process engineer would write on a notepad — and runs them continuously against live OPC-UA and historian data. We don't sell a data visualization layer that calls itself a digital twin. We don't build statistical surrogate models trained on your historian. We build mechanistic process models that can extrapolate to conditions you've never seen before, because the physics still apply.
Pittsburgh — a city built on process industries.
We didn't pick Pittsburgh by accident. The Pittsburgh metropolitan area has been home to continuous-process manufacturing for over a century — steel, coke, specialty chemicals, and now pharmaceutical contract manufacturing. That institutional knowledge is embedded in the engineering community here.
Our pilot plants are within driving distance of our office at 600 Grant Street. We can show up in person when the situation calls for it. That matters in process industries, where context is everything and a control room walkthrough tells you more than a SCADA screenshot ever will.
We're also embedded in the Carnegie Mellon / University of Pittsburgh engineering ecosystem, which has given us access to research collaborators in process systems engineering and model-predictive control.
Three principles we don't negotiate on.
Physics over statistics
A regression trained on last quarter's historian data will fail the moment your feedstock changes. First-principles models don't. Every Twynvex deployment runs governing equations, not curve fits.
Transparent models
Every parameter in the model is inspectable. You can see the reflux ratio assumption, the VLE method, the heat transfer coefficient. This is not a black box — it's the same process model your simulation engineers understand.
Plant-first design
We configure Twynvex to your specific unit operations before you deploy. We don't send you a login and a configuration wizard. We scope the model with your process team and validate it against your historian before going live.
Meet the engineering team.
Twynvex is a small, technical team. Everyone who works here has spent time in a plant, not just writing code about plants.