expertise — 05
Computational design that answers to physics.
Generative and parametric exploration, simulation-in-the-loop optimisation, custom tooling and data-informed design. Machines search wider option spaces than any team can; people decide what ships.
who this is for
Brutal constraints, measurable targets.
Engineering-led teams with hard performance targets — mass, stiffness, thermal margin — and no slack left in manual iteration. Products with packaging problems that fight every layout. Design organisations drowning in repetitive CAD work a script should be doing. This practice is for problems where the option space is too large for intuition and too consequential for guesswork.
what we do
The offerings.
Generative & Parametric Exploration
We encode the real constraints — package, structure, process, cost — and generate option spaces no manual process would reach, then curate with human judgement. Used for architecture studies, lattice and lightweighting work, and layout problems with brutal constraints. You walk away with a defensible best-of-space design and the parametric model that produced it.
Simulation-in-the-Loop Optimisation
Geometry evolved directly against structural, thermal or flow targets, with manufacturing constraints enforced so results are producible, not printable sculpture. You walk away with components measurably better on the metric that matters — mass, stiffness, thermal margin — and the evidence trail behind the shape.
Custom Design Tooling
When the off-the-shelf tool doesn't exist, we write it: configuration generators, CAD automation, review pipelines, data converters between your PLM and reality. Small software, large leverage. You walk away with tooling your team owns, documented and handed over — not a dependency on us.
Data-Informed Design
Fielded products generate the best brief for the next one. We instrument products (with consent and restraint), analyse usage and service data, and turn it into ranked design changes for the next revision. You walk away with a feedback pipeline and a rev-2 backlog based on how the product is actually used.
how it runs
Encode, search, verify, decide.
Every computational engagement starts by encoding constraints your engineers sign off on — that encoding is half the value. Searches run wide, results are verified in simulation and against manufacture with the engineering practice, and a person makes the call. Nothing ships because an algorithm liked it.
- Constraint model your engineers reviewed and own
- Best-of-space design with the evidence trail behind the shape
- Verified performance delta on the metric that mattered
- Parametric models and tooling handed over, documented
straight answers
Asked often.
Is this "AI design"?
It's computation used where it earns its place — search, optimisation, automation, and AI where AI is the right tool. Every result is tested against physics and manufacturability before it's shown to you.
Which problems suit generative methods?
Hard constraint sets with measurable objectives: packaging, structural lightweighting, lattices, layouts, product configuration. Not styling — form language stays a human judgement.
Do we keep the tools?
Yes. Models, scripts and pipelines are deliverables — documented, handed over, yours. Building you a dependency on us would be the opposite of the job.
works with
The wider search inside every loop.
Computational work sharpens engineering, widens industrial design exploration, and powers the real-time assets behind XR and configurators. Frequent duty in robotics and energy hardware.