KServe

KServe is a Kubernetes-native platform for serving machine learning models at production scale, providing a standard model-serving interface across multiple ML frameworks (TensorFlow, PyTorch, scikit-learn, XGBoost, and others) with built-in autoscaling, canary rollouts, and explainability.

It abstracts away framework-specific serving differences behind a common InferenceService API, so switching or mixing model frameworks doesn't mean rebuilding the serving infrastructure each time.

Part of the Kubeflow ecosystem; used by ML platform teams standardizing how models get deployed and served across an organization running multiple frameworks.