Simplifies and scales machine learning workflows on Kubernetes. Supports diverse ML needs and easy deployment.
Kubeflow is a machine learning toolkit that aims to simplify, port, and scale machine learning workflows on Kubernetes. It provides a straightforward way to deploy best-of-breed open-source systems for machine learning to various infrastructures using Kubernetes. With Kubeflow, users can create and manage interactive Jupyter notebooks, customize notebook deployments and compute resources to suit their data science needs, and experiment with workflows locally before deploying them to the cloud. Kubeflow also includes a custom TensorFlow training job operator for distributed TensorFlow training jobs, and supports model serving using TensorFlow Serving container. Additionally, Kubeflow integrates with Seldon Core, NVIDIA Triton Inference Server, and MLRun Serving for deploying machine learning models at scale. Users can take advantage of Kubeflow Pipelines for deploying and managing end-to-end machine learning workflows, allowing for rapid and reliable experimentation. Kubeflow is not limited to TensorFlow and also supports PyTorch, Apache MXNet, MPI, XGBoost, Chainer, and more, making it a versatile choice for diverse machine learning needs. The Kubeflow community is open and welcoming, providing opportunities to join discussions, attend community calls, and connect with other like-minded individuals and organizations.