SoftBox AI

Developed a web-based AI rendering platform that transforms CAD models into photorealistic product visuals using open-source diffusion models.

Researched and integrated ControlNets, IP Adapters, and LoRA training via ComfyUI to improve structural accuracy and output control in AI-generated imagery.

Implemented the system using Three.js for interactive 3D scene setup, with a frontend built in Next.js and a cloud-based AI inference backend deployed on RunPod using Docker.

The project was developed through a combination of self-directed experimentation and AI-assisted coding using Cursor, enabling rapid iteration and deeper architectural understanding.

Gained over 10,000 impressions on LinkedIn and attracted beta user interest from industry leaders at Amazon, Bentley, The North Face, and Logitech.