Best in class feature store, integrated into your MLOps stack
Open source. Fully supported. Built for scale.
Charmed Feast is Canonical's securely designed, production-ready feature store for machine learning. Deploy it in minutes alongside Charmed Kubeflow, and manage features across environments with confidence.
Install Charmed Feast Contact us
All the features of upstream Feast
Charmed Feast delivers all the features of upstream Feast that you know and love, including:
- Centralized feature registry
- Offline and online stores for consistent training and serving
- Real-time and batch feature ingestion
- Versioning and reproducibility for ML pipelines
Learn more about Feast
Trusted, supported, production-ready
Part of the Ubuntu Pro subscription
Regular updates and CVE patches
Optional 24/7 enterprise-grade support with SLAs
Operationalized and portable by design
Built as a Juju operator, Charmed Feast brings a new level of simplicity and flexibility to your feature store with easy deployment and scaling.
- Run it anywhere: cloud, hybrid, on-prem
- Integrate with Charmed Kubeflow out of the box
- Seamlessly plug into Feast from Charmed Kubeflow Jupyter Notebooks
Deploy with Juju
Use Charmed Feast with
Feast integrates natively with Charmed Kubeflow and other Canonical's MLOps tools.
Resources
Take your models to production using open source MLOps platforms.
Watch the webinar to learn how to build ML pipelines using Kubeflow.