Sunflower developed a first-of-it's-kind VST for realtime AI vocal transformation, using the latest AI voice technology. We designed and implemented the entire project, from design and conception to implementation, including the landing page and user area, the backend for payments and subscriptions, a custom realtime machine learning audio inference pipeline, as well as the actual application that runs locally on users' devices. We also researched new ML-audio models with a university collaborator, which involved implementing research papers from scratch in PyTorch, experimental runs, and mass data collection. This project involved technologies including PyTorch, React, C++ / JUCE, Node.JS, and GCP / Firebase, among others.