H2O eScorer
The external scoring tool that deploys a trained model's scoring artifact into a system you already run, such as Snowflake or Amazon SageMaker, with no new endpoint to operate.
- H2O eScorer, an add-on for deploying and scoring models outside H2O AI Cloud's own Kubernetes environment, runs a trained model inside a system you already operate, such as Snowflake or Amazon SageMaker.
- It scores MOJOs from H2O Driverless AI and H2O-3 Secure, and Python scoring pipelines, and returns Shapley contributions with each prediction.
- eScorer is a sibling to H2O MLOps, not a replacement: choose eScorer when the scoring must run inside an external system, MLOps when the model is served and monitored in H2O AI Cloud.
- eScorer is a paid add-on to H2O AI Cloud, installed per request.
- Cisco uses H2O AI Managed Cloud to build models and H2O eScorer to deploy them on the factory floor.
Deployment without a new endpoint to run.
H2O MLOps serves models in H2O AI Cloud. eScorer runs them inside a system your team already operates and governs.
- Snowflake scoringRuns MOJOs inside Snowflake as Java UDFs, container or external functions, or in Snowpark Container Services, so the data never leaves the warehouse.
- SageMaker endpointsPackages eScorer Standalone as a container behind an Amazon SageMaker real-time endpoint, for teams that standardize model serving on AWS.
- Database ScorerReads a table over JDBC or from S3, scores every row in batch, and writes the predictions back where your reports already read them.
- eScorer StandaloneA self-contained Java 17 server for on-premises hosts, remote plants, and edge sites, deployed and monitored from the eScorer UI.
- Many models per serverOne stateless eScorer instance hosts many MOJOs and Python scoring pipelines, so you run one service instead of a container per model.
- Reasons with every scoreReturns Shapley contributions with each prediction, in real time or written back in batch, so the reason code travels with the score.
From a trained model to a score in your system.
- 01Bring the modelUpload a MOJO from Driverless AI or H2O-3 Secure, or a Python scoring pipeline. Or import it from the H2O MLOps registry.
- 02Pick the targetChoose where the scoring runs: Snowflake, a SageMaker endpoint, a database job, a remote instance, or the eScorer service in H2O AI Cloud.
- 03Generate the codeAutogen writes the Snowflake functions, Jupyter notebooks, or batch properties files for the target, so nobody hand-writes deployment code per model.
- 04Score and watchCall it over REST, the Python client, or SQL. The Model Stats dashboard tracks invocations, latency, and errors for every deployment.
The sibling to MLOps, not a step in front of it.
eScorer and H2O MLOps are both ways to put a model into production. You pick one per use case by where the scoring has to run, not both in sequence. Highlighted: Observability, and MLOps, the tool eScorer sits beside.
What ships today
You operate it, or H2O.ai operates it for you.
Same platform either way: the same capabilities, the same governance, and the same audit trail. The choice stays open, so you can switch between them as the business changes, without rebuilding what runs on top.
H2O AI Cloud for Government
H2O AI Cloud for Government is FedRAMP High certified, delivering the highest standard of cloud security so federal agencies can harness predictive and generative AI without compromising data sovereignty or compliance.
Questions people ask before they call us.
H2O MLOps serves models in H2O AI Cloud and monitors them, with drift monitoring and alerts. eScorer runs the scoring artifact inside a system your team already operates, such as Snowflake or SageMaker. You pick one per use case by where the scoring has to run; they are not steps in one pipeline. eScorer can import models from the MLOps registry.
Bring the system where your scoring already has to happen.
We will show a scoring artifact deployed into Snowflake or SageMaker, side by side with an H2O MLOps endpoint.