Cloud Native Live: AI Inference with Crossplane & Modelplane

As open-weight models spread, platform teams are left running GPU capacity scattered across clouds, neoclouds, and on-premises clusters. Serving a model on one cluster is well covered by open source tools. The hard problems now sit above the cluster. In this Cloud Native Live session, we'll show how Crossplane handles that layer. We'll walk through Modelplane, an Apache 2.0 project built entirely as a Crossplane configuration, with no custom controllers. You'll see how composition functions written in Python, typed schemas from the new Crossplane CLI, and Crossplane v2 namespaced composites come together to schedule models across a multi-cluster fleet. We'll also cover the upstream fixes this work drove into Crossplane. We'll show it all through live demos of Modelplane and Crossplane, from both the platform team and ML team sides. This session is for platform engineers who want to see how far compositions and functions can go on a demanding real-world problem.