Our work spans the path from incoming requests to accelerators and distributed execution. We study how networking, systems software, and specialized hardware can reduce latency, improve throughput, and use computing resources more efficiently.
Areas of interest
Research directions include LLM inference and serving, semantic routing, cache and data movement, DPU and SmartNIC acceleration, high-performance networking, and distributed systems. These areas evolve as hardware and research questions change.
How we work
The lab combines measurement, prototyping, and systems experimentation across heterogeneous compute and network hardware. We build and evaluate end-to-end systems, from kernel and data-plane components to orchestration and model-serving software.