FriendliAI, founded in 2021, is a technology company spun out of Seoul National University research. It operates from Seoul, South Korea, and San Francisco, USA, focusing on AI/ML, enterprise software, and cloud infrastructure.
The company's core business is providing a fast AI inference platform for deploying and running large language models. Its technical work centers on AI inference optimization, GPU kernel development, caching, continuous batching, speculative decoding, and parallel inference. The FriendliAI Inference Platform claims to deliver over 2x faster inference and provides instant access to more than 590,000 models from Hugging Face.
FriendliAI offers enterprise-grade solutions including Dedicated Endpoints and container-based deployment options, both backed by a 99.99% uptime SLA. The company's stated mission is to make deploying large language models fast, affordable, and reliable.
