ADDITIVE-MANUFACTURING.ENGINEERING

In-house implicit engine · On-prem AI cluster

Additive manufacturing designs no traditional CAD could draw.

We generate, refine, and print organic, load-optimized geometry with a proprietary volumetric engine built on OpenVDB — driven by generative AI models running entirely on our own GPU cluster. No cloud. No compromise on complexity.

The Engine

A volumetric implicit engine, not a mesh editor

Every design starts as a signed distance field inside a sparse OpenVDB grid — not a triangle mesh. Boolean unions, TPMS lattices, graded infill, and organic blends happen natively at voxel resolution, without the topology errors that break traditional CAD kernels.

  • Sparse VDB grids — Millions of voxels, only where geometry exists — efficient enough to model full assemblies at print resolution.
  • Signed distance fields — Every surface is a mathematical field, so offsets, shells, and lattices are exact, not approximated.
  • Native TPMS & gyroid infill — Graded-density lattices generated directly in the field, tuned per region for stiffness, weight, or flow.
  • Print-ready by construction — Wall thickness, overhang angles, and self-supporting geometry are enforced in the field before a single triangle exists.

The Cluster

Generative AI, running entirely in-house

Our design models run on a private GPU cluster we own and operate — trained on manufacturing constraints, not stock imagery. Every part shown here was proposed, scored, and iterated by AI before a human ever opened a viewer.

  • On-prem, not outsourced — No data leaves our racks — full control over models, weights, and training data.
  • Constraint-aware generation — Models understand overhang limits, minimum wall thickness, and material behavior, not just aesthetics.
  • Thousands of variants per brief — Rapid generative exploration narrows a huge design space down to a shortlist worth printing.
  • Direct feedback into the engine — Every AI-proposed geometry is converted straight into the implicit field for refinement, never redrawn by hand.

Process

From prompt to printed part

1

Brief & constraints

Function, loads, material, and process constraints define the design space.

2

AI generative exploration

Our cluster proposes and scores thousands of geometric variants against the brief.

3

Implicit refinement

The best candidates are refined inside the OpenVDB engine: lattices, shells, blends, supports.

4

Print-ready output

Field data is sliced directly — no intermediate mesh cleanup — ready for the machine.

Have a part that deserves this?

Tell us about the geometry, the constraints, and the process. We'll show you what the cluster comes up with.