01 Write the physics
Governing equations, objective functions, material constraints. No sketches, no CAD. The specification is the only input FORGE needs.
VYO TECHNOLOGIES PRIVATE CANADIAN RESEARCH COMPANY
We compilephysics intohardware.
GPU-accelerated engines that turn governing equations into manufactured reality: generative geometry, rocket propulsion, tactical UAS, and the software that prints them.
00 THE PREMISE
You don't design the shape. You describe the physics. The shape was always there.
DESIGN IS NOT INVENTED. IT IS SOLVED FOR.
01 FLAGSHIP ENGINE
FORGE
FIELD-OPTIMIZED RIEMANNIAN GEOMETRY ENGINE
02 Flow on the manifold
The design space becomes a Riemannian manifold over a GPU voxel field. Structural, thermal, electromagnetic and fluid losses descend as one coupled problem, not in sequence.
03 Read the geometry
The field converges to the shape that could not have been any other shape. Holes appear, channels merge, topology changes, without mesh surgery. Export straight to manufacturing.
FIELD STATEENTROPY
∂L/∂G → 0ITER 0000
One manifold, every domain
Thermal, structural, EM and fluid objectives co-optimized simultaneously; a change in one resolves through all.
OpenVDB voxel kernel
Industrial-grade sparse voxel geometry, PicoGK-compatible .vdb export into the LEAP 71 ecosystem.
CUDA-accelerated descent
Riemannian gradient flow evaluated directly on the GPU field: millions of voxels per step.
Manufacturing-aware
Minimum feature size is encoded in the field itself; the same representation drives the printer.
02 PROPULSION
AGNI अग्नि
AGNI does not design rocket engines. It grows them, from the interaction of physics, constraints and geometry on a curved space.
Give it propellant, thrust class and chamber pressure. Geometric diffusion on a Riemannian manifold grows the chamber, regenerative cooling channels, injector and turbomachinery as a single coupled field. What comes back is print-ready geometry, fully documented.
0min
Spec → print-ready geometry
0
Subsystems, one unified field
0
Hand-drawn surfaces
DEONTIC CONSTRAINTS · BIFURCATION-CONTROLLED ANNEALING · GPU STENCIL COMPUTE
03 ADDITIVE MANUFACTURING
The slicing
engine.
A Rust-built precision slicer for resin (MSLA) printing: mesh to machine file, with a cryptographic paper trail. It feeds a parallel photocuring system that exposes layers on multiple light engines at once.
- Deterministic provenance
Every slice emits a SHA-256 manifest: supply-chain attestation built into the build file itself.
- Engineered in Rust
A workspace of focused crates behind one engine facade, exercised by ~800 tests.
- Generative supports
Pillar, compound and lattice supports computed from the geometry, not hand-placed.
- Parallel photocuring
Layers allocated round-robin across 3-4 linear light engines on a rotating platform, seams feather-blended.
0+
Unit & integration tests
0
Mesh formats ingested
0×
Throughput, parallel curing
LAYER DECOMPOSITION01 / 16
50 µmSEAM-BLENDED · 4 LIGHT ENGINES
04 DEFENCE & AUTONOMY
S.U.R.G.E.
SCALABLE UAS RAPID GENERATION ENGINE
The whole stack, pointed at the sky: tactical unmanned aerial systems engineered, optimized and manufactured at surge rate, in Canada.
S.U.R.G.E. turns the engines behind FORGE, AGNI and the slicing platform toward one mission: taking tactical UAS from specification to fielded fleet without the bottlenecks of conventional tooling. Developed as a participant in the Department of National Defence IDEaS Drone Surge challenge.
- Surge-rate production
Additive-first pipelines designed to scale from a single airframe to a fleet without retooling.
- Stack-native design
Airframes and propulsion co-optimized by FORGE-class physics compilation, not adapted from commercial platforms.
- Attestable supply chain
Every printed component carries the slicing engine’s SHA-256 provenance manifest: traceability by default.
DND IDEAS · DRONE SURGE PARTICIPANT ◆ TACTICAL UAS ◆ CANADIAN SOVEREIGN MANUFACTURING
PRODUCTION RUNQUEUED
BATCH 0481 → 1 UNITS · SURGE RATE
05 CORE TECHNOLOGY
One stack, four
disciplines.
GPU-accelerated computing
CUDA stencil kernels, sparse voxel fields and parallel field evolution: millions of degrees of freedom per optimization step.
CUDA · OPENVDB · NANOVDBRiemannian optimization
Design spaces as curved manifolds. Constraints become curvature; the optimizer flows along geodesics instead of searching.
GEOMETRIC DIFFUSION · LANGEVIN DYNAMICSAdditive manufacturing
From slicing kernels and generative supports to multi-light-engine photocuring control: software down to the photon.
MSLA · LPBF · NON-PLANAR PATHSAutonomous & space systems
Trajectory planning, satellite coverage and robotic platforms for defense and space, built on the same compute core.
SATPATH · ROBOTICS · DRONES06 IN THE LAB
Research that hasn't
shipped yet.
A rotating bench of prototypes feeding the core stack. When one matures, it graduates into the product line above.
SatPath
Trajectory and coverage planning for satellite constellations and space-based platforms.
● ACTIVESIMCAD
GPU-native parametric CAD in Rust: sketches, manifolds and AI-assisted modeling on one kernel.
● ACTIVENon-planar toolpaths
Curved-layer deposition planning with A*, RRT and BVH collision checking: printing beyond flat slices.
● PROTOTYPESIM-LANG
A domain language for composing physical simulations the way you compose software.
● PROTOTYPE07 WORK WITH US
Build the inevitable with us.
We partner with research groups, manufacturers and mission teams who need hardware their current tools can't reach. Tell us the physics. We'll find the shape.
PRIVATE CANADIAN RESEARCH COMPANY · ENGINEERED IN CANADA