I am a 1st-year PhD student in the Computer System Lab of ECE Department at Cornell University. Before joining Cornell, I received my B.S. in Applied Physics from Peking University (EECS Department) in 2026, advised by Prof. Meng Li. During my undergraduate studies, I spent a year as a research intern in the SHARC Lab at Georgia Tech, working with Prof. Cong (Callie) Hao on efficient Mixture-of-Agents (MoA) serving systems.

Research interests: optimizing large-scale GPU systems for more efficient machine learning systems.

1. Efficient agentic systems — achieving lower serving latency via co-designing scheduling policy and universal memory management. Related work: Faster-MoA (DAC’26).
2. Hardware-aware system/algorithms design — designing tailored system/algorithms on resource-constrained platforms to enable higher performance. Related work: CREATE (ASPLOS’26) and DySL-VLA (DAC’26).

Hands-on experience

• Efficient embodied agents: fault-tolerant embodied AI (CREATE) and robotic-manipulation acceleration (DySL-VLA).
• Algorithms: PEFT fine-tuning of small models with LLaMA-Factory; customized metrics for dynamic early exit.
• Systems: KV-cache management and transmission; prefill–decode scheduling and partitioning in multi-agent serving.
• Hardware: sparse matrix-matrix multiplication (SpMM) accelerator and systolic-array design in Verilog.