Hi, I'm Lipenghee 👋

PhD candidate at Harbin Institute of Technology, working on efficient inference for large language and vision-language models.

Research Interests

• Quantization and mixed-precision inference
• Adaptive and inference-time computation
• Multimodal model efficiency
• Model-system co-design

Current Focus

• Efficient inference algorithms for LLMs and VLMs
• Reproducible evaluation of accuracy, memory, latency, and throughput
• Open-source contributions to modern LLM inference systems

Technical Stack

Python · PyTorch · Transformers · vLLM · Linux · Git

Open Source

Building public contributions around multimodal inference, KV-cache efficiency, and reliable performance evaluation.