I am an assistant professor in the Department of Computer Science at Duke University. Before Duke, I was a postdoctoral researcher at UC Berkeley, working with Professor Ion Stoica. I completed my Ph.D. in the Paul G. Allen School of Computer Science and Engineering at University of Washington, advised by Professor Tom Anderson and Professor Arvind Krishnamurthy. My research has received NSF CAREER Award (2023), USENIX Security Distinguished Paper Award (2023), USENIX FAST Best Paper Award (2021), Amazon Research Award (2021), Google Academic Research Award (2024), IBM Academic Award (2021), and Meta Research Award (2022, 2021).
My research interests are (1) datacenter and cloud computing and (2) machine learning systems. Some research hightlights:
Phoenix, a system service that exposes common application-level abstractions for speed and manageability. Its components include:
MCCS (SIGCOMM24), an collective communication system that decouples algorithm choice from applications, allowing the system to adjust algorithm choice without application interruption.
mRPC (NSDI23), an RPC architecture that supports efficient RPC-level policies and live upgradability.
RDMABench, a test framework for RDMA software/hardware stack. Its components include:
Husky (NSDI23), a test suite for RDMA performance isolation for multi-tenancy. Our follow-up work, Harmonic (NSDI24), is the first step towards the performance isolation issues found by Husky.
Collie (NSDI22), a test suite for RDMA performance anomalies.
A series of machine learning system research, including:
VTC (OSDI24), a large language model serving system with fairness guarantees.
Punica (MLSys24), a serving system for LoRA fine-tuned large language models.