H2O Driverless AI
The leading AutoML platform that automates feature engineering, model selection, and tuning, and documents every model as it trains.
- H2O Driverless AI (DAI) is an AutoML platform that automates six stages: AutoViz, feature engineering, model tuning, AutoDoc, AutoMLI, and scoring pipelines.
- AutoDoc has shipped in every Driverless AI release since 1.2 in June 2018. AutoDoc and AutoMLI are capabilities of the product, not add-ons.
- Driverless AI engineers features by evolutionary search over transformations, interactions, and encodings, then searches and ensembles models on GPU.
- Commonwealth Bank of Australia cut customer scam losses 76% from their 2022 peak with H2O predictive models, protecting more than 16 million customers.
Six things it automates.
Each one was a bottleneck for a data science team. Each one is now a checkbox.
- AutoVizAutomatic visualization of the dataset before modeling: distributions, outliers, correlations, and missing values. See the problem before you model it.
- Feature engineeringEvolutionary search over transformations, interactions, and encodings. The feature engineering that separates a Grandmaster from a good analyst, run automatically.
- Model tuningAlgorithm selection, hyperparameter search, and ensembling on GPU, with time and accuracy dials you control instead of a black box.
- AutoDocModel documentation written as the model trains: data, features, methods, validation, and results, in the format a model risk team expects.
- AutoMLIK-LIME, Shapley, surrogate decision trees, partial dependence, disparate impact analysis, and a reason code for every single prediction.
- Scoring pipelinesPython pipelines for batch and low-latency Java MOJO for real time, deployable to MLOps, Hadoop, Spark, Teradata, or Oracle.
From dataset to documented model.
- 01Connect dataFiles, JDBC, Snowflake, Databricks, S3, or Hadoop. Or pull ready features straight from the Feature Store.
- 02Set the dialsAccuracy, time, interpretability. Choose recipes, pick a target, and let AutoViz show you the data first.
- 03Run the experimentFeature engineering, model search, and ensembling run on GPU. Watch the leaderboard form as it goes.
- 04Ship with the paperworkAutoDoc and AutoMLI come out with the model. Export a MOJO, register in MLOps, or hand it to the Data Science Agent.
The training engine at the center of the predictive lane.
Fed by the Feature Store. Served through MLOps. Driven by the Data Science Agent. Highlighted is Driverless AI.
What ships today
The Data Science Agent drives it. It does not replace it.
Ask the H2O AI Super Agent to build a model and its Data Science Agent plans the experiment, runs Driverless AI, reads the leaderboard, and returns the AutoDoc. The engine that has cleared model risk reviews since 2017 is what the agent operates.
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
The H2O AI Cloud for Government has achieved FedRAMP High Certification, delivering the highest standard of cloud security so federal agencies can harness predictive and generative AI without compromising data sovereignty or compliance.
H2O Driverless AI in production
:quality(75))
:quality(75))
:quality(75))
Watch H2O Driverless AI in action
Questions people ask before they call us.
Driverless AI automates six stages. They are AutoViz (data visualization), feature engineering by evolutionary search, model selection and hyperparameter tuning with ensembling on GPU, AutoDoc (model documentation), AutoMLI (interpretability: K-LIME, Shapley, surrogate trees, disparate impact, reason codes), and scoring pipelines in Python and Java MOJO.
Bring your hardest tabular problem and last quarter's data.
We will run it through Driverless AI with you and hand your model risk team the AutoDoc the same day.