RadixArk is an infrastructure-focused company building large-scale inference and training systems for the AI community. The company originated from research at UC Berkeley's lab of Ion Stoica, co-founder of Databricks, and maintains an open-source-first approach to its products.
The company's primary offering is SGLang, an open-source inference engine for serving modern AI models. Launched in 2023 from UC Berkeley, SGLang has accumulated over 20,000 GitHub stars and gained adoption from companies including xAI and Cursor. RadixArk also develops Miles, an open-source project for reinforcement learning training. Both products form the foundation of the company's broader mission to make frontier-level AI infrastructure accessible to developers, startups, enterprises, and research labs.
RadixArk is valued at approximately $400 million and has raised funding led by Accel, with angel investment from Lip-Bu Tan. The company's leadership brings experience from AI-focused technology firms and research environments, with CEO and co-founder Ying Sheng having previously worked at xAI and as a research scientist at Databricks.