Curriculum Vitae

Education

Zhejiang University · Master’s in Software Engineering
School of Software Technology · Sep 2025 – Jun 2028 (expected)

Harbin Institute of Technology · B.Eng. in Intelligent Vehicle Engineering
School of Automotive Engineering · Sep 2021 – Jun 2025

Research

AWARe: Mitigating Catastrophic Forgetting via Activation-Weighted Adaptive REtaining

• Studied catastrophic forgetting in multimodal large language models.
• Estimated parameter importance from activation distributions and retained high-importance weights during continual learning.

Experience

Uni-Ubi · Algorithm Intern · Apr 2025 – Sep 2025

• Fine-tuned Qwen2.5-VL with LLaMA Factory for construction-site safety inspection.
• Trained models with VeRL and designed LLM-as-a-judge reward functions.
• Worked with vLLM and validated the NVFP4 quantization workflow.

Open-source Contributions

• vLLM: added FP8 GEMM kernel support for NVIDIA Blackwell GPUs (PR #17280).
• SGLang: added FP8 GEMM kernel support for NVIDIA Blackwell GPUs (PR #9403).
• VeRL: added video-training support using vLLM for rollout (PR #3928).
• Transformers: added GLM-4.1V training support (PR #39199).

Skills

• Programming: Python, C/C++, CUDA, multithreading, and network programming
• Machine learning: PyTorch, Transformers, multimodal LLMs, reinforcement learning, model quantization
• Systems: Linux, Git, Docker, vLLM, distributed training with FSDP and Ray
• Robotics: ROS 2, NVIDIA Jetson, STM32, and embedded systems