Hello! I’m an undergraduate researcher at the EMIL Lab, Kyung Hee University, advised by Professor Sungha Choi. My research focuses on building efficient algorithms and systems that make multimodal foundation models practical on real-world devices, even under tight memory, compute, and latency constraints.
Efficient Multimodal Adaptation on Edge Devices
Integrated a camera streaming pipeline into the SDK to support humanoid locomotion.
Investigate the intersection of emotional factors and learning outcomes through a multidisciplinary approach, in collaboration with Dept. of Metaverse, fostering global academic partnerships.
Developing modality-aware KV cache quantization for long-video MLLMs, exploiting temporal redundancy across video tokens and distinct characteristics of visual and text KV caches. The goal is to enable memory-efficient long-video understanding while preserving model performance.