Kyoung-Min Kim

Kyoung-Min Kim is a Researcher at the Data-Intensive Applications and Systems Laboratory (DIAS) within the College of Computing and Communication Sciences at École Polytechnique Fédérale de Lausanne (EPFL). His work intersects large language models (LLMs) and database systems, focusing on improving inference efficiency and reliability.

• Email: kyoung-min.kim@epfl.ch
• Office: BC 224, EPFL, Station 14, Lausanne

Research Interests: Recent publications highlight his expertise in:

• LLM inference optimization
• GPU resource management
• Cost modeling for AI systems
• Scheduling algorithms for scalable inference
• Preemption strategies reducing computational costs
• Constraint satisfaction frameworks for system performance

Key Contributions: Kim co-developed INFERMAX, an analytical framework comparing LLM inference schedulers. His work demonstrates 30% GPU cost reduction through preemption techniques.