What ZenML does
ZenML is an open-source orchestration platform that unifies machine learning and generative AI workflows. It acts as a control plane and metadata layer that abstracts infrastructure, letting teams write pipeline code once and run it across different clouds and environments.
Key capabilities
Data scientists build pipelines using Python decorators, with support for frameworks like PyTorch, TensorFlow, scikit-learn, and LLMs. ZenML handles containerization, GPU provisioning, and scaling across local, Kubernetes, AWS, or GCP setups. It automatically versions artifacts and models for reproducibility and rollback, and includes a runtime for durable execution of AI agents with crash recovery and checkpointing.
Who it's for
ZenML serves engineering teams building production ML and LLM systems, including organizations moving from notebooks to scalable pipelines and teams needing dev-to-production parity. The open-source core is complemented by managed features for governance, SSO, and centralized control.