Struct

generate any 3D asset to manufacturing

Comment

Project Struct

Any 3D asset → Manufacturing CAD

Inspiration

Access to building physical products is still limited.
CAD tools are complex, expensive, and require years of training, while AI tools generate outputs that are not editable or usable for real engineering.

We asked:

How can we use AI to make designing real-world objects accessible to anyone?

Struct explores this by turning raw 3D inputs into editable, structured CAD workflows, lowering the barrier to engineering and manufacturing.

What it does

Struct converts raw 3D inputs into an interactive CAD workflow in the browser.

  • Upload a 3D file (.stl, .glb, .ply)
  • Or generate one from text/image
  • Convert it into CadQuery (parametric code)
  • Edit the code directly
  • Rebuild into STEP (CAD format)
  • View instantly in 3D

** input → code → edit → rebuild → visualize **

Social Impact (UN SDGs)

🏭 SDG 9 — Industry, Innovation & Infrastructure

  • Lowers barriers to manufacturing tools
  • Enables faster prototyping and innovation

🎓 SDG 4 — Quality Education

  • Makes CAD and engineering more accessible
  • Allows learning through direct interaction with models

💼 SDG 8 — Decent Work & Economic Growth

  • Empowers makers, students, and startups to build products
  • Reduces cost and time to enter hardware development

♻️ SDG 12 — Responsible Production

  • Encourages efficient iteration and reduces wasted prototypes

How I built it

Frontend

  • Next.js, React, TypeScript
  • Three.js for 3D visualization

Backend

  • FastAPI
  • CadQuery (OCCT)

Pipeline

  1. Input (upload / generate)
  2. Backend → CadQuery code (async)
  3. Edit in UI
  4. Execute → STEP
  5. Parse + render in browser

Challenges

  • Converting mesh → structured CAD
  • Handling async generation + polling
  • Executing CAD code safely
  • Parsing STEP in the browser
  • Keeping interaction fast and simple

Accomplishments

  • Built full mesh → code → CAD viewer pipeline
  • Enabled live editing of CAD in-browser
  • Integrated generation + editing into one workflow
  • Delivered an end-to-end prototype

What I learned

  • Mesh generation alone is not enough for real-world use
  • Code provides a powerful interface for geometry
  • The hardest problem is translating user intent into structure
  • Accessibility is as important as capability

What's next

  • Improve mesh → CAD reconstruction
  • Add smarter parametric understanding
  • Support more formats (DXF, IGES)
  • Add manufacturability (DFM) checks
  • Move toward a deterministic CAD pipeline

Summary

Project Struct explores one idea:

Using AI to make designing real-world objects accessible, editable, and usable — for anyone.

Built With

Share this project:

Updates