Text2CAD is a neural network-based system that generates procedural CAD designs from text descriptions at varying expertise levels. The approach uses transformer architectures (BERT-based) trained on sequential CAD data to map text inputs to parametric CAD sequences representing 3D geometries. Published at NeurIPS 2024 as a Spotlight paper, the project includes data preparation pipelines, training code, and inference utilities for the text-to-3D generation task.