About

Hi there and welcome! I'm Wenyao Gao. I build and optimize systems for serving large language models — inference engines, KV cache management, batching, and the GPU infrastructure underneath.

My work involves designing high-performance inference pipelines, understanding the intricacies of GPU memory management, and implementing efficient batching strategies. I'm particularly interested in frameworks like vLLM and SGLang that push the boundaries of what's possible in LLM serving.