Serving platforms and GitOps: KServe, KubeRay, llm-d, Helm and Argo
When raw Deployments stop being enough: what KServe, KubeRay, llm-d and Dynamo-style platforms add, how to choose, packaging with Helm and Kustomize, GitOps with Argo CD or Flux, progressive delivery, and a decision framework that favours the simplest thing that works.
By now you can write the Deployment, Service, PDB and HPA for a model by hand. The question this lesson answers is what to do when that stops being enough, and how to keep the YAML itself under control.
A caution that applies to this entire area: the serving-platform ecosystem changes quickly. Project names, APIs and recommended patterns shift between releases. Treat what follows as a map of concepts and trade-offs, and check each project's current documentation before you commit.
Pro content
Keep reading with Inference Academy Pro
Unlock the rest of this lesson, plus every advanced lesson, hands-on lab, the larger playground models and completion certificates.