Case study — 2025
SolidMake
Role
AI Research · Engineering
Stack
Custom AI · CAD
In one line
Describe the part. Get the STEP file.
A purpose-built text-to-CAD system for mechanical engineers — not a GPT wrapper. Describe the part in plain language; get a production-ready STEP file for cutting, bending and welding.
01 — The problem
Mechanical design time is dominated by modeling — hours in SolidWorks or AutoCAD for parts an engineer can describe in a sentence. The bottleneck isn't imagination; it's geometry entry.
02 — The model
SolidMake builds on DeepCAD and Text2CAD, assisted by prompt normalization and pre-designed training so natural language reliably becomes up-to-the-mark geometry. The process flow itself was optimized end to end — this is custom model work, not API glue.
03 — The output
Designs export as STEP files — the format fabrication actually runs on. From a sentence to something you can cut, bend and weld.
04 — The result
My final-year project, and the deepest technical build in this portfolio — the proof behind the words 'custom AI models'.
Next project
AI SEO Automator
SEO that runs itself.