The cookie-cutter template for Kubeflow deployment.
This template generates a Terraform script that can deploy a Kubeflow production-ready deployment.
It is set up to use GCS and CloudSQL as backends to the service with automatically deployed Kubeflow and Kubeflow Serving into GCP without requiring an AI platform. It provides an optional secure endpoint for Kubeflow Serving (TLS and IAP) and secures an attempt to optimize GKE resource usage and thus billing costs.
Quick start on GCP with a ready-to-use code and project
IAM: manage roles and permissions on user and domain level, or access with service account
Manage multiple environments (dev, testing, staging, etc.), trace activity and changes
Handle blueprint versions
Automatic PR update on new version (security patch)
Least privilege principle security (list of needed permissions to be able to provision it per blueprint)
Provisioning: infrastructure as a code included
The Kubeflow blueprint includes industry best practices in the form of ready-to-use codes or templates delivered to you, on your infrastructure, through seamless deployment for sustainable quality.
The platform allows you to standardize workflows. Developing the models with maximum efficiency will not interfere with integration and will manage the risk of a lower ROI due to productizing and integration.
Applying best practices in production requires continual improvement and support. Leveraging on the blueprint repository of Aliz Intelligence Platform ensures that the technology is always up to date.
By using one-click deployment of a Kubeflow cluster, you can customize it to your needs, and then simply deploy everything with Terraform.
By leveraging Google Cloud Platform's managed solutions, your pipeline artifacts are always safe.