I am an incoming graduate student at the Music and Audio Computing Lab (MACLab), KAIST. I study how music can be represented, understood, and generated—especially through audio tokenization, semantic representations, and post-training.
My goal is to build music AI systems that are not only capable, but also controllable, musically coherent, and genuinely useful to creators.
| Work | Venue | What it explores |
|---|---|---|
| DuoTok | ACM MM 2026 · Oral | Source-aware dual-track tokenization for vocal–accompaniment generation |
| Back to Ear | INTERSPEECH 2026 · Accepted | Perceptually driven high-fidelity music reconstruction |
- Preparing to join KAIST MACLab in Fall 2026
- Exploring post-training and preference learning for music models
- Building semantic music representations and creator-facing generation tools
Music Generation · Audio Tokenization · Music Understanding · Post-training · Preference Learning
Continuous sound → discrete representation → creative possibility