AI-Driven Computational Geometry — Merging ChatGPT with PicoGK

This project introduces an AI-powered geometry design tool that fuses the creative reasoning of ChatGPT with the computational precision of PicoGK. The system was led and fully implemented by Eric Lehderas part of an internal R&D effort at Pipeline Organics.

You can try the prototype here (still in development — please avoid clicking "Generate STL" for now):
https://black-river-048387203.1.azurestaticapps.net/

AI Geometry Tool UI

1. Concept — A New Way to Design Geometry

Designed and implemented by Eric Lehder, the tool replaces manual scripting of lattices, implicit fields, and voxel logic with a natural-language interface. Engineers describe what they want, and the AI generates a complete PicoGK-compatible C# file that follows strict templates and geometric conventions.

This creates a conversational workflow for algorithmic geometry: the engineer specifies intent, the AI writes the code, and PicoGK executes it — merging creativity and computational accuracy into a single seamless process.

2. Architecture — ChatGPT, PicoGK, and Real-Time Feedback

The platform is structured around three interactive panes: the AI Designer, the 3D Preview, and the Result. Together, they form a fast and iterative computational geometry pipeline.

  • AI Designer: Accepts natural language prompts such as "generate a TPMS electrode with variable pore size".
  • Backend Compiler: Writes the produced code to disk, compiles it, and runs PicoGK to generate STL outputs.
  • 3D Preview: Displays the resulting STL as soon as the geometry is generated (this feature is being finalised).
  • Result Panel: Shows AI reasoning, explanations, or parametric suggestions.

Compilation errors automatically feed back into the AI, creating a self-correcting iterative loop. The system effectively blends deterministic geometry generation with adaptive AI reasoning.

3. Applications — From Electrodes to Lattices

Developed by Eric Lehder to support rapid R&D cycles, the tool can generate a wide range of implicit and voxel-based geometries:

  • TPMS lattices with tunable pore size and thickness
  • Spiral and radial electrode architectures
  • Multi-material arrangements using voxel fusion
  • Cylindrical and organic implicit surfaces
  • Beam networks and structural scaffolds
  • Tank-arc inspired geometries and concept shapes

This accelerates early-stage hardware development by allowing engineers to explore design spaces in seconds, rather than writing thousands of lines of code manually.

Summary

By combining ChatGPT’s generative power with PicoGK’s computational modelling backend, this tool — led and implemented by Eric Lehder — introduces a new paradigm for geometry creation. It enables rapid exploration of complex structures for additive manufacturing, energy systems, and advanced computational design, all through the simplicity of natural language.