ZenML

ZenML is an MLOps framework designed to facilitate infrastructure-agnostic machine learning pipelines. It offers a managed control plane known as ZenML Pro and provides over 50 integrations to ease workflow management. ZenML supports a variety of solutions including Generative AI (GenAI) and Large Language Models (LLMs), fine-tuning LLMs, and creating coding copilots. The framework allows for training on Spot VMs and deploying Sagemaker Endpoints, as well as managing GCP Vertex AI and training on Kubernetes. ZenML also includes features for creating alerting and custom stack components, and it offers a fully flexible framework that can integrate with standard MLOps tools or custom components. Additionally, ZenML provides a VSCode extension to work directly with ZenML pipelines and stacks from the IDE. The company offers a cloud product that includes a managed version of the open-source framework with additional cloud-only features and supports multiple deployment scenarios, including self-hosted and managed SaaS. ZenML provides a central model registry for ML models, allowing for deployment, monitoring, and updating from a single dashboard. It also offers role-based access control (RBAC) to secure pipelines and stacks and supports tenant-based deployments for isolating work across multiple projects and teams. The open-source framework is available on GitHub and can be managed in-house without using the cloud product. ZenML’s cloud product offers enhanced observability with features like a model control plane and CI/CD capabilities. Project templates help in building ML models faster by providing a structured codebase and commonly-used libraries. The new open-source dashboard unifies the experience for both OSS and cloud users. ZenML’s adoption by companies like Brevo and ADEO Leroy Merlin has significantly improved their ML deployment efficiency and reduced time-to-market.