H2O Driverless AI

What it does

H2O Driverless AI takes in tabular datasets, ingested from sources such as Hadoop HDFS and Amazon S3, and a user-defined target column. It automates feature engineering, model development, validation, tuning, and selection, then produces scoring pipelines exported as Python modules and Java standalone artifacts, plus model documentation and interpretability reports.

People use it to move from raw data to a deployable model with less manual work. Its distinct elements are automatic feature engineering, a guidance wizard that recommends settings based on the dataset and business requirements, and machine learning interpretability at both global and individual-prediction levels. It is designed to use GPU acceleration on NVIDIA hardware and runs on commodity hardware.

Driverless AI suits data scientists, DevOps teams deploying models, and business analysts who need predictive models, particularly in regulated industries where explainability matters. It is an installed platform configured through documentation, so access depends on an installation or licensed environment. Limitations include reliance on hardware for GPU speedups and a feature set documented for a specific version line.

What it can do

• Automatic feature engineering that detects relevant features and interactions in a dataset
• Automated model development, validation, tuning, and selection
• Machine learning interpretability with global and local explanations
• AI Wizard that recommends settings based on data and business requirements
• Automated model documentation (AutoDoc)
• Deployment as REST endpoint, cloud service, or Java scoring code

Works with

• Hadoop HDFS
• Amazon S3
• Snowflake
• H2O AI Cloud workspace
• Triton Inference Server
• H2O MLOps

Example workflows

Build a predictive model from a tabular dataset

Illustrative example. Input: a dataset in HDFS or S3 and a target column. Steps: import the data, let automatic visualization and feature engineering run, let automated model development and validation select the algorithm, review interpretability output. Output: a selected model with validation results and an exported scoring pipeline.

Deploy a model for production scoring

Illustrative example. Input: a validated model from an experiment. Steps: choose a deployment target such as a REST endpoint, a cloud service, or Java scoring code for edge devices. Output: a running endpoint or exported scoring artifact.

Before you choose

• Publicly stated GPU speedups apply to supported NVIDIA hardware and specific algorithms such as XGBoost and LightGBM.
• This profile describes Driverless AI specifically, not H2O AI Cloud or the h2oGPTe generative AI product, which the documentation references separately.

Listed under

• Data
• Analytics and BI
• Automl
• Data science
• Enterprise
• Feature engineering
• Machine learning
• Model explainability

Get started

Demo request is the primary entry route on the product page; installation documentation covers deployment.

1. Review the installation documentation and choose Docker, native, or cloud installation.
2. Configure licensing and authentication as described in the configuration documentation.
3. Import a dataset, define the target column, and start an experiment, optionally using the AI Wizard.

Driverless AI requires an appropriate license and compute environment. Confirm license cost, trial eligibility and deployment requirements with H2O.ai.

We couldn’t find published prices for H2O Driverless AI. See their pricing page ↗

Sources for this overview

Product pages, documentation, and pricing checked 29 Oct 2026. Each source supports part of the overview; availability and prices can change.

• H2O Driverless AI ↗h2oai.com
• Overview — Using Driverless AI 2.5.1 documentation ↗docs.h2o.ai

Company website

h2oai.com: product, pricing and policy pages

YC directory mirror

Batch and company name

Web search

Confirms the official site

Social accounts

Accounts the company links, and its posts on X

Human review

Checked by a person

knowledge_base

Last checked 29 Oct 2026. Every field keeps its source and date; unknown means we couldn’t confirm it.