RadixArk
Democratizing frontier AI infrastructure
About
RadixArk is a commercial spin-out from UC Berkeley launched in January 2025 following the open-source release of SGLang at Berkeley SkyLab in 2024. The company develops scalable AI inference infrastructure, positioning SGLang as a core inference engine integrated with leading AI platforms like xAI. RadixArk raised $100 million in seed funding led by Accel at a $400 million valuation to expand access to high-performance AI compute for frontier models. The startup’s business model focuses on delivering advanced AI compute infrastructure solutions that accelerate and optimize AI deployments for AI-first enterprises and research institutions.
Products
-
SGLang
— Fast, flexible open engine for serving modern AI models with high performance and reliability in production.
Customer sectors: AI infrastructure engineers, enterprise AI teams, research labs
-
Miles
— Open-source framework for large-scale reinforcement learning training, bringing rigor and scalability to RL workflows.
Customer sectors: AI researchers, reinforcement learning engineers, AI startups
-
Managed infrastructure and tooling
— Cloud-hosted infrastructure and tools enabling developers and enterprises to build, train, and run AI systems at scale.
Customer sectors: AI developers, startups, enterprises
Market research & competitive positioning
Source-backed TAM / SAM / SOM sizing, displacement analysis, and an interactive peer positioning map for RadixArk — available to members.
Related Companies
Other startups in Developer Tools & Infrastructure
