My research focuses on making computing and networking infrastructure predictable and resilient for diverse applications, including AI training and inference and real-time interactive services, across cloud datacenters and the network edge. I build abstractions and runtime systems that deliver predictable performance and resilience under resource sharing and failures, and develop methods to predict and validate system behavior through efficient performance modeling and realistic testing.
Recent News [all]
- [09/2026] NicOS and Programmable Scale-Up Switches accepted to appear at HotNets 2026
- [09/2026] Dooly accepted to appear at NeurIPS 2026
- [08/2026] Vulcan accepted to appear at EuroSys 2027
- [07/2026] Received an NSF VINES award for real-time agricultural robotics with AI-native cellular edge computing
- [07/2026] LFS accepted to appear at NSDI 2027
- [05/2026] PacketExpress accepted to appear at SIGCOMM 2026
- [03/2026] Mimesys accepted to appear at OSDI 2026
- [12/2025] SMEC accepted to appear at NSDI 2026
Research Themes
Programmable and Resilient Datacenter Infrastructure for AI and Cloud
Abstractions and runtime systems for programming heterogeneous hardware (SmartNICs, switches, and GPU fabrics), managing shared compute, memory, and network resources, and recovering efficiently from failures, including in large-scale AI training
Leto (arXiv'26) · NicOS (HotNets'26) · Programmable Scale-Up Switches (HotNets'26) · Alkali (NSDI'25) · ExoPlane (NSDI'23) · RedPlane (SIGCOMM'21) · TEA (SIGCOMM'20)
Predictable and Resilient Real-Time Edge Computing
Systems and abstractions for sharing wireless network and nearby server resources to deliver low-latency, real-time interactive applications, from live video analytics to live generative AI, with predictable performance, security, and resilience
SMEC (NSDI'26) · ARMA (MobiSys'25) · 5G Fronthaul Security (USENIX Security'24) · Slingshot (SIGCOMM'23) · Atlas (MobiCom'23)
Predicting and Validating System Behavior
Low-cost performance models and realistic, executable workloads for predicting and validating how systems behave across hardware, configurations, and resource contention, without exhaustive profiling or access to production systems
Dooly (NeurIPS'26) · Mimesys (OSDI'26) · TraceLLM (EMNLP'25)