Research

In-Network Acceleration

the emergence of programmable network devices (P4 switches, SmartNICs, and DPUs more recently) has motivated a new concept called in-network computing. We are exploring how programmable network elements can better contribute to computation acceleration in the post-Moore's law era.

Sustainable IoT

IoT systems are ubiquitous nowadays. Yet, virtually all IoT devices rely on battery to function. Batteries are hazardous, prone to disasters, and hard to maintain --- all indicating that current IoT systems are not sustainable. We are working towards sustainable IoT systems by removing completely the batteries from IoT devices.

Systems Support for Machine Learning

Machine learning is powering many of our intellegent services in our daily life. Yet, serving computation-intensive machine learning workloads while meeting their stringent throughput/latency requirements remains a critical challenge. We are developing efficient network systems to support machine learning inference.