AI-native CAD
with a dual-engine kernel.

NeuroCAD combines an SDF geometry engine with a Geometric Algebra engine, synchronized through five graph layers for browser-native mechanical design.

NeuroCAD — CAD platform interface
13 Rust Crates Modular kernel architecture.
30+ Field Operators Compose, evolve, optimize.
WASM WASM-Native Zero install. Runs in the browser.
EU EU-Based Bulgarian company. GDPR-native.
PLATFORM CAPABILITIES

Built as a platform.
Not a plugin.

From parametric sketching to simulation-ready geometry. One product surface over two mathematical engines and a coherent graph stack.

Sketch & Constraint Authoring

Deterministic constraint solver with canonical outcomes. Lines, arcs, circles, splines with geometric and dimensional constraints.

Feature Kernel

Implemented kernel paths for sketch-to-extrude and sketch-to-revolve, plus feature history and edit-intent infrastructure. Fillet/chamfer and richer feature workflows are being hardened for pilots.

Surface Modeling

Surface layer with NURBS admission/evaluation, loft/sweep execution contracts, surface session workflows, and quality/healing diagnostics. UI-level workflows are being productized.

Organic & Field Modeling

Field-native geometry with PDE-driven shape evolution. Sculpt, deform, and evolve surfaces with exact-zone protection.

Feature-Preserving Booleans

SDF booleans and smooth variants are implemented in kernel/session paths. Feature-preserving boolean reports and evidence exist, with integration hardening ongoing.

Diffusion & Evolution

Reaction-diffusion systems, mean curvature flow, Laplace-Beltrami operators. Bounded field evolution with replay safety.

Differentiable CAD

Gradient exposure for lawful geometry optimization. Sensitivities flow through the entire design graph.

Lattice Engineering

Graded lattice generation with parametric unit cells. Gyroid, diamond, octet — tuned for strength, weight, or thermal conductivity.

Material Fields

Continuous material grading from one zone to another. Per-voxel multi-material assignment for PolyJet and MJF printing.

Multi-Scale Composition

Macro structure + micro lattice + material grading, coordinated as a single design. Scales from millimeters to meters.

Manufacturing Governance

Manufacturing constraints are moving into governed design workflows. Wall thickness, draft, overhang, export, and validation evidence are tracked through deterministic gates.

Metrology & Inspection

Measurement, tolerancing, GD&T, QIF/DMIS, and inspection evidence loops for validation-driven engineering workflows.

NeuroCAD — Real kernel geometry
Real kernel geometry in the browser
NeuroCAD — Multiple primitives
Multiple primitives with parametric tree
NeuroCAD — Sketch to 3D extrusion
Sketch extruded to 3D solid
TECHNOLOGY

Two mathematical engines.
One coherent CAD state.

DUAL MATHEMATICAL KERNEL

  • Engine A: Signed Distance Fields and nalgebra/f64 evaluation for numeric geometry
  • Engine B: Geometric Algebra in CGA + PGA for structure, transforms, and motors
  • Dual-engine coherence checks compare Engine B projections against Engine A values
  • E-graphs provide equality saturation and canonical-form extraction
  • DCG, BVH, and NeuroGraph keep updates coherent across committed states

EXECUTION

  • WASM compilation — kernel runs in the browser
  • WebGPU — GPU-accelerated field evaluation and rendering
  • Web Workers — kernel in background, UI always responsive
  • Deterministic replay — every design reproducible from graph
  • CRDT collaboration — real-time multi-user editing

MATHEMATICS

  • PDE operators — reaction-diffusion, curvature flow, Laplace-Beltrami
  • Adaptive field resolution — multi-scale refinement
  • Differentiable geometry — gradients for optimization
  • Variational reconstruction — surface fitting from point clouds

INTEGRATION

  • MOOSE bridge — finite element simulation coupling
  • Multi-format export — STL, STEP with semantic overlays
  • AI orchestration boundary — proposals validated, never source of truth
DUAL ENGINE ARCHITECTURE
SDF + GA, synchronized by graphs

Not only SDF. NeuroCAD has two complementary mathematical engines.

The REV11.2 specification defines Engine A as SDF/nalgebra numeric evaluation and Engine B as Geometric Algebra in CGA and PGA. The five graph layers keep these engines synchronized in a committed CAD state.

ENGINE A

SDF numeric geometry

Signed Distance Fields and nalgebra/f64 evaluation provide the numeric geometry truth: inside/outside, distance, gradients, booleans, offsets, and field composition.

ENGINE B

Geometric Algebra CGA + PGA

CGA represents spheres, planes, circles, and tangent constructions. PGA represents rigid motion, screw axes, motors, mates, and mechanism paths.

COHERENCE

Projection and audit gates

Overlap operations are reconciled through projection π and per-operation tolerance ε. Engine disagreement is treated as a correctness signal, not hidden tolerance magic.

ENGINE A / SDF ENGINE B / CGA + PGA π / ε COHERENCE GATE
kernel_graph operation history and canonical NodeId allocator
DCG dependency tracking and dirty propagation
e-graph equality saturation and canonical-form extraction
BVH spatial index for fast field queries
NeuroGraph documentary segmentation and history layer
SIMULATION AND PHYSICS AI

Meet NeuroFEM.

A companion path for physics-aware design.

NeuroFEM is the simulation and physics-intelligence direction around NeuroCAD. The immediate product goal is credible pilot workflows: geometry creation, validation, simulation handoff, and evidence-driven iteration.

Simulation-Ready Geometry

SDF-native geometry is designed to remain analyzable. The platform direction is to reduce the gap between authoring, meshing, export, and FEM validation.

Physics Model Roadmap

The roadmap includes a foundation physics model and GPU training infrastructure. This is treated as an evidence-building program, not a claim of finished production accuracy.

Autonomous Engineering Factory

The same autonomous software factory used to build NeuroCAD is being upgraded with serious server and GPU capacity so engineering execution can move faster with auditability.

HOW WE COMPARE

Not another CAD.
A different foundation.

Capability NeuroCAD SolidWorks Fusion 360 nTop Onshape
Implicit Geometry (SDF)
Parametric Sketching
Surface Modeling
Lattice Engineering
Differentiable CAD
Browser-Native ~
WebGPU
Deterministic Replay
Material Fields
Manufacturing Governance ~ ~
AI Orchestration ~
Simulation Bridge ~
USE CASES

From aerospace to medical.
One platform.

Additive Manufacturing

  • Per-voxel multi-material assignment
  • Lattice grading for weight optimization
  • Conformal cooling channel design
  • 4D printing of stimulus-responsive structures
  • Functional surface texturing

Aerospace

  • Structural topology optimization
  • Multi-scale composite design
  • Acoustic metamaterial panels
  • Thermal management geometry
  • Lightweight lattice wings

Medical Devices

  • Bio-inspired trabecular bone structures
  • Conformal fit from CT/MRI scans
  • Functional biocompatible surfaces
  • Multi-material gradient implants
  • Scan-to-CAD reconstruction

Automotive

  • NVH optimization with acoustic metamaterials
  • Conformal cooling for injection molding
  • Lightweight suspension components
  • Topology-optimized brackets
  • Manufacturing-aware geometry

Consumer Products

  • Organic form generation
  • Programmable mechanical properties
  • Topology optimization workflow setup
  • Procedural texturing for 3D printing
  • Multi-material soft/rigid zones

Energy & Sustainability

  • Heat exchanger optimization
  • Lattice structures for conductivity
  • Solar panel self-cleaning surfaces
  • Wind turbine blade internal lattice
  • Thermal energy harvesting devices
ROADMAP

Built in phases.
Validated with pilots.

R1 is the foundation layer. The investor story is the broader platform: pilot geometry, design intelligence, and industrial validation are already represented as tested kernel workflows under product hardening.

R1 — WORKING FOUNDATION

Core Geometry MVP

MVP LIVE
  • Sketching and constraints
  • SDF primitives and booleans
  • Interactive field preview and meshing
  • STL export path
  • Deterministic replay
R2 — PILOT GEOMETRY

Feature & Surface Kernel

PILOT FOCUS
  • Sketch extrude/revolve are live SDF features
  • NURBS, loft, and sweep kernel paths are implemented
  • Healing diagnostics and validation gates are in place
  • Smooth SDF booleans are live, with feature-preserving evidence layers
  • Surface-field conversion is moving through pilot hardening
R3 — DESIGN INTELLIGENCE

AI-Assisted Design & Optimization Workflows

TESTED FLOWS
  • Organic field modeling with protected exact-zone gates
  • Gated diffusion / neural design operations
  • Differentiable CAD signals for gradients, curvature, and SDF deltas
  • Material fields with deterministic geometry coupling
  • Multi-scale geometry composition across smooth SDF, morph/warp, and lattice fields
R4 — INDUSTRIALIZATION

Manufacturing & Validation

EVIDENCE GATES
  • Lattice engineering workflows with certification gates
  • Metrology, GD&T, QIF/DMIS, and inspection evidence loops
  • Manufacturing governance, BOM/PDM, and export readiness gates
  • Topology optimization workflow scaffold under hardening
  • TPMS, strut, Voronoi, plate, and neural lattice pattern libraries
PILOT ACCESS

Design partner access, not public SaaS pricing.

NeuroCAD is moving from MVP/demo toward focused pilot and design partner validation.

Demo Review

Intro /current path

Short technical and business review for investors, advisors, and potential design partners.

  • Working MVP/demo overview
  • Architecture walkthrough
  • Current status split
  • Pilot fit discussion
  • Follow-up materials
Request Access

Enterprise / Strategic

Custom /by discussion

For industrial teams, strategic partners, and investors evaluating deeper technical collaboration.

  • Technical due-diligence session
  • Roadmap alignment
  • Security/IP discussion
  • Pilot or investment process
  • Legal review when needed
Start Discussion

Discuss NeuroCAD as a pilot or investment opportunity.

Working MVP/demo available for qualified investor and design-partner conversations.