Balancing hypergrowth with optimization

H2O.ai is a visionary leader in democratizing AI and it has built a SaaS on AWS. They needed a way to improve the provisioning time for their tenants, who use H2O.ai's cloud platform to run machine learning models.

H2O.ai used *Karpenter*, a Kubernetes-native cluster autoscaler, to automatically scale their Kubernetes clusters up and down based on demand. They also used Bottlerocket OS, a lightweight, secure Linux distribution designed for containers. And they used prefetching to cachecontainer images so that they could be provisioned more quickly.

As a result of these changes, H2O.ai was able to improvetheir tenant provisioning time by *100x*. This allowed them to scale their business more quickly and meet the needs of their growing customer base.