Multi-modal LLM Serving
Efficient and predictable serving for mixed multimodal LLM workloads.
About
I am a Postdoctoral Researcher at KAIST CASYS. My research focuses on systems and architectures for LLM inference and serving, with interests in multimodal workloads, KV-cache management, and processing-in-memory. I received my Ph.D. in Computer Science from KAIST, advised by Prof. Jaehyuk Huh.
Research Focus
Hierarchical KV Cache Management
Memory-efficient LLM serving through hierarchical KV-cache management.
Processing-in-Memory for LLM Serving
System and architecture support for efficient PIM-based LLM serving.
Education
Ph.D. in Computer Science, KAIST
Dissertation: SLO-aware Inference Serving Systems for Multi-Modal Generative-AI Workloads
M.S. in Computer Science, KAIST
B.S. in Computer Science, KAIST
Publications
Loading publications...
Contact
Email: leesh6796@casys.kaist.ac.kr
Location: KAIST, Daejeon, South Korea