김영재 연구실

System-level optimizations for RAG-based LLM serving including heterogeneous resource utilization, data movement minimization, KV cache management, and QoS-aware adaptive scheduling for latency-stable inference.

Software optimization for distributed deep learning including Graph Neural Networks (GNN), model and data parallelism, and efficient resource utilization for vision, language, and graph models.

Vector database architecture for billion-scale datasets with storage-based ANN search, index partitioning, and stream processing integration for RAG and recommendation systems.

LSM-tree optimization addressing write amplification, read overhead, and space amplification with hardware acceleration for AI and big data workloads including KV cache management.

CXL-based memory expansion and optimization for deep learning models, GNN-based recommendation systems, and disaggregated computing environments with improved memory and I/O efficiency.

Object storage design for unstructured data management in data lakes, cloud-native applications, and machine learning with metadata-driven retrieval and flexible data organization.

In-memory streaming platform optimization for heterogeneous computing infrastructures including query execution plans, task scheduling, and state management for real-time data processing.