News

June 29, 2026 Introducing Ekka: automated diagnosis of silent errors in LLM inference via differential debugging, accepted to ICML 2026!

June 25, 2026 Introducing TraceLab: real-world coding-agent traces for systems research and optimization, with an open pipeline to collect, analyze, and replay your own traces.

June 19, 2026 Introducing M*: a modular, extensible serving system that treats multimodal models as graphs and requests as Walks

May 28, 2026 Team UW SyFI wins three awards at the FlashInfer AI Kernel Generation Contest at MLSys 2026!

May 12, 2026 Introducing VibeServe: let AI agents write your LLM serving stack end-to-end