I’m a first-year PhD student at Carnegie Mellon University, advised by Prof. Shinji Watanabe. Before joining CMU, I earned my B.S. in Computer Science and Engineering from Seoul National University and worked at Samsung Research.

My background is in speech processing, particularly speech recognition and spoken keyword spotting. At Samsung Research, I worked on deploying and optimizing speech technologies for Samsung Bixby, covering large-scale training, model compression, long-form speech recognition, and cross-modal inference-time biasing.

My current research focuses on speech-to-text technologies:

  • Fully open-source speech-to-text models: OWSM is an academic reproduction of OpenAI Whisper that aims to build speech foundation models that are fully open, from the training framework to the datasets. In that spirit, I’m developing data optimization techniques, including improved CTC alignment and methods for measuring dataset contamination.

  • Multimodal representation learning: Beyond my current work, I hope to extend my research in the following directions:

    • Few-/zero-shot alignment of multimodal representations by exploiting the isomorphism between them.
    • LM-agnostic speech encoders that align layers of the speech encoder with the corresponding layers of the underlying large language model according to their representation granularity.


News!

:loudspeaker: (08/24/2026) I am joining WAVLab @ LTI/CMU!

:loudspeaker: (09/24/2024) I will be presenting my paper on spoken keyword detection at INTERSPEECH 2024!