Home / Technology / From Berkeley Lab to $400M AI Startup
From Berkeley Lab to $400M AI Startup
22 Jan
Summary
- SGLang developers formed RadixArk, valued at $400 million.
- RadixArk focuses on optimizing AI inference processing.
- The startup is developing Miles for reinforcement learning.
A significant portion of the team behind SGLang, a widely-used open-source tool for AI model training acceleration, has transitioned to a new commercial venture named RadixArk. This startup, announced last August, has reportedly achieved a valuation of about $400 million following a funding round led by Accel. RadixArk emerged from UC Berkeley's lab in 2023.
Ying Sheng, a primary contributor to SGLang and former engineer at xAI, is now the co-founder and CEO of RadixArk. The startup, like SGLang, concentrates on optimizing inference processing, a critical component for enhancing AI model speed and efficiency on existing hardware.
Alongside its continued development of SGLang, RadixArk is also creating Miles, a dedicated framework for reinforcement learning. This aims to empower businesses to progressively improve their AI models. While SGLang remains free, RadixArk has introduced fees for its hosting services, reflecting a growing trend in the AI inference infrastructure sector.
