Biography
I am a first-year Ph.D. student in Computer Science at Johns Hopkins University. I am interested in how understanding the behavior of distributed AI workloads can help us build more efficient and adaptive computing infrastructure.
My long-term vision is to enable people to independently choose, customize, and deploy AI. Toward that vision, I want to unlock the potential of available computing resources through workload understanding and system optimization.
My current focus is mechanism-aware communication optimization for distributed AI. I investigate how GPU collective operations and their surrounding execution phases produce network traffic bursts, then aim to use these insights to design better scheduling, routing, and load-balancing mechanisms. Before JHU, I worked on proportional-fair multipath forwarding and network traffic prediction, and was a research intern at the Korea Institute of Science and Technology (KIST) studying multimodal spatiotemporal data.
Study workloads
Connect execution phases and GPU collective operations to observed communication patterns.
Locate bottlenecks
Use cross-layer experiments to distinguish collective, NIC, and network effects behind traffic bursts.
Design mechanisms
Explore adaptive scheduling, routing, and load balancing for efficient distributed AI.
Selected Projects
All projects ↗Understanding Traffic Bursts in GPU Collective Communication
Mechanism-aware, cross-layer measurement of GPU collective traffic bursts and the execution phases that shape them.
Proportional-Fair Multipath Forwarding in Edge Networks
Designed and tested traffic-aware multipath forwarding to distribute data across constrained network paths.