From CLI to Creative UX: Democratizing AI Access for Artists

The core philosophy behind this project was accessibility. In the current AI landscape, groundbreaking models are often "trapped" within complex command-line interfaces. I realized that if a tool requires a 20-minute environment setup, it’s effectively broken for 90% of potential users. While scientists and engineers are comfortable with environment setup, dependency management, and CUDA configurations, these technical barriers prevent creative professionals from exploring the model's potential.

For generative models like MusicLDM, direct interaction with artists is indispensable. Artists can provide the essential intuition and user feedback to the researchers that will drive the model’s evolution.

My goal was to gain hands-on experience in transforming complex AI systems into inclusive platforms where tools are no longer exclusive to those who can navigate a terminal. To achieve this, I developed a functional prototype and Proof of Concept (PoC) on the UCSD DSMLP (Data Science Machine Learning Platform). This project serves as a bridge to connect remote GPU clusters with intuitive, web-based access for anyone with a creative vision.

Note: The platform was hosted and accessible for the duration of the Spring Quarter 2024 on UCSD DSMLP.