About ODAHU

The Open Data AI Hub (ODAHU) is an open source project that provides the building blocks for enterprise grade MLOps platforms.

Multi ML Frameworks

• Supporting major ML frameworks: Scikit-learn, Keras, Tensorflow, PyTorch, H2O (and more)
• Extends MLflow services with enterprise level features

Multi Clouds

• Kubernetes native system of services
• Deployment automation to Kubernetes cluster with Helm charts
• Supporting major Kubernetes platforms: AWS EKS, Azure AKS, GCP GKE, RedHat OpenShift

Secure

• Single sign-on (SSO) based on OAuth2
• RESTful API secured with SSL
• Role based access control based on Open Policy Agent
• Users activity audit
• Credentials and keys manager based on HashiCorp Vault
• Internal traffic encryption with Istio

Modular and Extensible

• Services for different ML phases: transform, train, validate, package, deploy, evaluate
• Services are extensible and manageable via REST APIs, SDK and CLI
• Functionality extensible with new services
• Connectors for data sources, package repositories, Docker container registries
• Plugins for data science IDEs
• Plugins for workflow engines like Airflow

Scalable

• Systems based on ODAHU components can be scaled from small to very large.
• Scalable ML model training, packaging, serving components

Manageable

• Pre-build monitoring dashboards
• Configurable alerting rules
• Configurable logs collection
• Compatible with third party logs processing systems

Open

• It is free and open-source with the Apache2 License.
• Contribution to project is welcome!