시각 및 지능 시스템 설계 연구실

SYDLAB envisions the future of computing for artificial intelligence. As large-scale AI models continue to grow in scale, memory systems—not computation—have become the primary performance bottleneck. We develop memory-centric architectures, digital circuit designs, and next-generation AI accelerators, along with hardware-software co-designed systems that enable efficient execution of LLMs, multimodal AI, and distributed agent systems. We aim to redefine how intelligent systems are built and deployed by bridging the gap between hardware and AI, enabling scalable, adaptive, and efficient computing platforms that power the next generation of intelligent machines across cloud, edge, and physical environments. We design next-generation AI memory systems using CXL, HBM, and heterogeneous memory architectures to accelerate large language models and data-intensive AI workloads. We design NPU, GPU, FPGA, and ASIC accelerators for efficient AI inference, focusing on high-performance digital architectures and domain-specific hardware. We accelerate AI models using quantization, pruning, speculative decoding, sparsity, and GPU-aware optimization for efficient LLM and vision model deployment. We explore next-generation AI computer systems integrating CPUs, GPUs, NPUs, CXL memory, and high-bandwidth interconnects. We develop multi-agent AI systems that communicate, coordinate, and reason efficiently using shared memory, distributed execution, and collaborative intelligence. We build next-generation visual AI systems using Gaussian representations, multimodal models, and efficient video understanding for robotics and Physical AI. From 2002 to 2005, she was with the Digital TV Development Group, Samsung Electronics Company Ltd., Suwon, South Korea, as an Engineer, where she was involv