KAIST Startup Village needed an environment where startups on campus could efficiently use shared GPU resources and reliably conduct the AI R&D each company needs. It also required a framework that provides machine learning and deep learning development environments together with high-performance computing, to support resident companies' AI technology development and commercialization.

The key challenge was to let many resident companies efficiently share limited GPU resources while continuing research and experiments in their own independent AI development environments.

With Cheetah, we provided an operating framework to manage resident companies in groups and flexibly allocate GPU resources, along with built-in AI development environments and multi-node cluster capabilities.

KAIST Startup Village now manages 15 resident companies in groups, with an environment for distributing and using shared GPU resources more systematically. Resident companies can set up independent AI development environments with their allocated resources and handle complex AI workloads more efficiently with multi-node clusters.