SkyPilot - Run AI on Any Cloud

Discover how SkyPilot, an open-source framework, enables running LLMs, AI, and batch jobs on any cloud in this 30-minute talk. Learn about SkyPilot's innovative approach to viewing all clouds and regions as a coherent collection called the Sky, offering AI practitioners high GPU availability, maximum cost savings, and managed execution through a simple cloud-agnostic interface. Explore how SkyPilot transparently sends jobs to the most cost-effective and available locations, manages execution, and reduces cloud costs by supporting spot instances with automatic recovery and cleanup of idle cloud instances. Gain insights from real-world deployment experiences, including its use in popular LLM projects like Vicuna chatbots and the vLLM inference system, as well as its adoption by numerous organizations for diverse use cases such as AI training on GPUs/TPUs, AI serving, CPU batch jobs, and interactive development.