Song-Ze Yu

I'm Song-Ze, a CS student split between UC Berkeley and National Tsing Hua University, doing research at BAIR and CNMAT. At BAIR, I'm building Anamnesis, a platform for large-scale AI-powered survey simulation. At CNMAT, I'm working on audio-language models that let you edit sound through natural language. I also lead the ML Pod at Berkeleytime, which serves 122K+ monthly users. Outside research, I'm a musician who once improvised live songs from audience prompts at NCKU. Ask me about audio-language models, LLM persona simulation, semantic search, and music tech.