CV

Education

• B.S. in Computer Science, Zhejiang University, 2022 - 2026(Expected)

Research experience

UCSD Picasso Lab | Mar 2025 – present
Advisor: Yufei Ding
Optimize KV-cache and LoRA usage for multi-agent workflows under the SGLang framework; modified SGLang backend engine, cache structures, write queues, interrupt modules, and LoRA scheduler.

Zhejiang University CCNT Lab | June 2024 – Nov 2025
Advisor: Peng Lin
Designed a high-accuracy, low-power neural network combining LIF spike features and LSTM gating for lab neuromorphic hardware.

Zhejiang University RC4ML Lab | Nov 2024 – Mar 2025
Advisor: Zeke Wang
Accelerated convolution with large-number multiplication algorithms on DSP48.
Evaluated AMD VERSAL Network-on-Chip performance on Vivado.

Publications

KVFlow: Efficient Prefix Caching for Accelerating LLM-Based Multi-Agent Workflows Permalink

Scaling Multi-Agent Simulation

Unlocking Long-Term Dependencies in Spiking Neural Networks with a Recurrent LIF Memory Module

Skills

• Football
  • I used to play football for ZJU-CCST team.
• Guitar
  • I was the president of Zhejiang University Guitar Association
  • I can play acoustic guitar, electronic guitar, and a little Bass / Violin
• Photography
  • Lumix S5 with 50f1.8, 70300f5.6