Beyond Human CAD. EvoPhase Uses AI to Strip the Bias from Fluid Geometry.
04 Nov 2026 · Stories from AIMX by Industrial Gemba’s Editorial Team EvoPhase Co-Founder and CEO Dominik Werner on optimising equipment geometry, designs that don’t resemble their originals,...
04 Nov 2026 · Stories from AIMX by Industrial Gemba’s Editorial Team
Quick Skim
EvoPhase Co-Founder and CEO Dominik Werner on optimising equipment geometry, designs that don’t resemble their originals, and a first trade show that surprised the team.
“As long as there’s something flowing around your equipment, we can optimise it,” says Dominik Werner, Co-Founder and CEO of EvoPhase.
The industrial engineering consultancy uses machine learning to redesign equipment geometry from the fluid dynamics up. Targeting fluid and powder handling across pharmaceuticals, food and beverage, and specialty chemicals, EvoPhase’s objective is direct: maximise manufacturing throughput and cut energy consumption by breaking away from conventional mechanical blueprints.
The Shape Is the Variable
The core technical differentiator, according to Dominik, lies in how the platform models geometry. By using a compact mathematical representation, the AI can manipulate surfaces and flow paths without human-imposed geometric bias.
That freedom produces radical results.
“In the end, you get equipment that might not look like a ribbon mixer, for example, or like a wind turbine,” Dominik explains. “But it functions like one—and it’s much more efficient.”
Unfamiliar Shapes, Plant-Floor Realities
While generative fluid design can yield impressive computational models, alien geometries inevitably face plant-floor scrutiny:
- Hard Metrics Over Simulation: Dominik’s pitch promises efficiency leaps, but industrial buyers need concrete deltas on batch times and power draw before altering a validated production line.
- Manufacturability & Hygiene: In regulated environments like pharma and food processing, wetted parts cannot simply perform well in fluid dynamic models. They must be machinable or 3D-printable without prohibitive cost, free of dead zones where bio-burden accumulates, and simple to clean in place (CIP). An algorithmically perfect shape is useless if maintenance cannot swab it.
Demand From Unexpected Sectors
AIMX marked EvoPhase’s debut on the trade show circuit, drawing attention well beyond the team’s initial target markets. Dominik described the reaction as “astounding,” noting strong interest from process engineers across sectors the consultancy had not originally factored into its roadmap.
The Gemba Takeaway
Before commissioning an algorithmic equipment redesign:
- Request Empirical Proof: Ask for measured, empirical before-and-after data on a production unit, not computational fluid dynamics (CFD) heatmaps. Focus strictly on throughput per kilowatt-hour.
- Audit Cleanability and Tooling: Bring your fabrication and validation leads into the first design review. Confirm early whether local machinists can hold tolerances on the organic contours, and verify that the design won’t fail sanitary cleaning audits.
This interview is part of Industrial Gemba’s coverage from AIMX – the region’s AI event featuring real-world use cases, innovation showcases, and strategic insights. Responses have been edited for length and clarity. Watch the full interview on YouTube.

