Friendli

FriendliAI operates a frontier AI inference cloud platform that enables enterprises to deploy open-weight and custom AI models with unmatched efficiency and cost-effectiveness. Its purpose-built inference stack delivers 2x+ faster performance through custom GPU kernels, smart caching, continuous batching, speculative decoding, and parallel inference, combined with infrastructure-level optimizations including multi-cloud scaling and geo-distributed deployment. The platform supports instant one-click deployment of over 590,000 Hugging Face models across language, audio, and vision modalities, and offers 99.99% uptime SLAs with enterprise-grade fault tolerance. FriendliAI provides Model APIs, Dedicated Endpoints, and container solutions, enabling customers to scale to trillions of tokens with up to 50% fewer GPUs. The company has formed strategic alliances with Samsung Cloud Platform and NVIDIA. Founded in 2021 and headquartered in Redwood City, California with a hub in Seoul, Korea, the company has raised approximately $20 million at the seed stage.