RadixArk

RadixArk is an infrastructure-first AI company building large-scale inference and training systems for the entire AI community. Founded by Ying Sheng (CEO, ex-xAI) and Banghua Zhu (CTO, UC Berkeley), RadixArk spun off from the SGLang open-source project to make frontier-level AI infrastructure open and accessible to everyone. For inference, the company builds on SGLang — the fastest open engine for serving modern models. For RL training, it develops Miles, an open-source framework for large-scale post-training. On top of these open cores, RadixArk ships managed infrastructure and tooling for individual developers, startups, enterprises, and research labs. The company launched publicly in May 2026 with $100M in seed funding at a $400M valuation, led by Accel.