Operate.

Create a central place to host and manage all experiments across your entire organization, enabling greater sharing and collaboration. Maintain a view of all deployment versions with complete model registry and model versioning management capabilities that are accessible by both an easy-to-use web interface and an API. You can also manage models trained on any 3rd party framework natively, and/or import models stored on a 3rd party model registry.

Build once and deploy to any scoring environment with target deployments.  Deploy your models in mode of your choice, including Single Model, A/B Testing, or Champion/Challenger. Your models can also be scored in real-time (synchronous or asynchronous), in batch, or as streaming data. Choose the Kubernetes configuration that is best suited for each model, including, CPU vs. GPU and min and max processing and memory allocations.

Enable high availability for your deployments by selecting up to 5 replicas for your deployment. H2O MLOps will automatically check the health of each node, and load balance across nodes, so if one node fails, customers won’t experience any service interruption.

Maintain oversight for your entire deployment, including model drift, model accuracy, and deployment infrastructure. Model explanations at runtime delivers local explanations as to which features are contributing the most or least to prediction values. You can set custom thresholds to receive alerts and notifications for all monitored metrics.