radixark
RadixArk is an infrastructure-first, deep-tech company building large-scale inference and training systems to democratize frontier AI infrastructure.
At a Glance
- AI model developers
- Enterprise IT
- Hyperscalers
- Open-source AI community
AI Tools by radixark
(1)
Miles
Open Source LLM RL Framework
Discussions
No discussions yet
Be the first to start a discussion about radixark
Latest News
RadixArk Launches with $100M in Seed Funding Led by Accel to Democratize AI Infrastructure
RadixArk raises $100 million to make AI chips more efficient
DeepSeek-V4 on Day 0: Fast Inference and Verified RL with SGLang and Miles
Andrew Ng and RadixArk Partner for 'Efficient LLM Inference with SGLang' Course
Products & Services
A high-performance open-source inference engine for large language models, featuring RadixAttention and structured generation.
An enterprise-facing reinforcement learning (RL) framework optimized for large-scale model post-training and production workloads.
A managed cloud platform providing optimized hosting and enterprise infrastructure for AI training and inference.
Market Position
RadixArk positions itself as a faster, more efficient alternative to standard inference engines, leveraging its academic roots and open-source dominance in structured generation.
Leadership
Founders
Ying Sheng
Co-founder and CEO. Former engineer at xAI, researcher at Databricks, and Ph.D. student at UC Berkeley. Key contributor to SGLang.
Banghua Zhu
Co-founder and CTO. Former engineer at NVIDIA and researcher at UC Berkeley. Expert in AI infrastructure and modeling.
Executive Team
Ying Sheng
CEO
Former xAI and Databricks engineer; UC Berkeley SGLang core contributor.
Banghua Zhu
CTO
Former NVIDIA engineer; AI infrastructure veteran.
Board of Directors
Founding Story
RadixArk was founded to commercialize SGLang, an open-source project incubated in Ion Stoica's lab at UC Berkeley. The founders, Ying Sheng and Banghua Zhu, aimed to solve the 'memory hog' problem of AI computing and make inference and training more efficient for the entire community.
Business Model
Revenue Model
Open-core model: Free open-source tools (SGLang, Miles) combined with paid hosting services, enterprise support, and managed infrastructure.
Pricing Tiers
Full access to SGLang and Miles core frameworks on GitHub.
Initial $200 credits offered to early supporters; fees for managed hosting and infrastructure.
Target Markets
- AI model developers
- Enterprise IT
- Hyperscalers
- Open-source AI community
- High-performance LLM serving
- Reinforcement learning post-training
- Large-scale MoE production workloads
- Edge and local hardware optimization
- DeepLearning.AI
- LMSYS Org
- Open-source developers using SGLang