I'm currently focusing on SGLang Serving, working at Alibaba Cloud โ๏ธ.
SGLang Maintainer โ mainly working on the KVCache Scheduling System of SGLang, covering hierarchical caching, sparse, PD disaggregation and serving systems for agentic workloads.
Roadmaps I'm Driving
- SGLang Distributed KVCache System For Agentic Workload
- SGLang's new lightweight SessionAware Router module for bucket-based PD serving
- HiSparse Roadmap
- Unified Radix Cache โ LMSYS Blog, 2026-08
- SGLang HiSparse โ LMSYS Blog, 2026-04
- HiSparse: Scaling Sparse-Attention Decoding with Hierarchical KV Cache Management โ Arvix, 2026-08