GatwAI

H2O Driverless AI

Understand H2O Driverless AI: automated feature engineering, forecasting, model interpretation, deployment, commercial licensing and practical limits.

Provider sources checked

What does H2O Driverless AI do?

H2O Driverless AI automates parts of machine learning: exploring data, generating features, selecting models and tuning parameters. Tasks include classification, regression and time-series forecasting. It targets teams developing repeatable predictive models from their own data, rather than users seeking a general knowledge chatbot. AutoViz assists exploration, AutoDoc records experiments and interpretation tools help inspect models. Scoring can be integrated through Python or Java/MOJO. Custom recipes extend the process. This flexibility creates responsibilities: preparing data, designing evaluation and handing models over reliably to operations remain the team's work.
Editorial research · 2026-11-14How we research →

Who is this tool for?

  • Data scientists and machine learning engineers.
  • Organizations with their own infrastructure and recurring forecasting tasks.

Strengths

  • Automated feature engineering and model comparison.
  • Documentation, interpretation and exportable scoring.
  • On-premises, air-gapped or private-cloud deployment options.

Limitations and requirements

  • Requires appropriate compute resources and a license.
  • Leakage or unsuitable samples can produce misleading test scores.
  • Model explanations do not establish causation.

Who is behind the product?

Company
H2O.ai
Headquarters
US vendor; current legal headquarters not separately checked here.
Company founded
Founding year not confirmed in the sources read for this profile.
Product launch
Driverless AI 1.0 was released in September 2017.

Driverless AI is a commercial product within H2O.ai's broader portfolio and should not be confused with open-source H2O-3.

Suggested ways to use it

Suggested workflows or verification steps. These are not hands-on tests; for unverified products they are not a recommendation to use the service.

Prepare inventory forecasting

  1. Check historical volumes, timestamps and permitted features for completeness.
  2. Evaluate a time-series experiment against a seasonal baseline.
  3. Inspect errors by item group and unusual weeks; document assumptions.

Goal: A proposed traceable model decision rather than simply the best aggregate metric.

Hand a model to operations

  1. Archive the selected experiment, AutoDoc and data version.
  2. Export its scoring artifact and compare predictions with the experiment.
  3. Define data-change thresholds and a fallback to the previous method.

Goal: An inspectable handover; GatwAI has not executed this workflow.

Pricing, allowances and data

What should you check before choosing a plan?

Commercial licenses and enterprise prices are arranged with sales. Documentation distinguishes production, development and evaluation licenses. A tutorial describes a 21-day trial key and a time-limited Aquarium lab. Your infrastructure costs and trial-key restrictions require separate checking.

What should you check about data and permissions?

Keeping data and models in your environment distinguishes the product from a purely SaaS chatbot. Access, deletion and security controls are still required. In particular, inspect network access by custom recipes, data connectors and additional models. Local installation does not automatically make every extension free of external communication.

External reviews and available evidence

Martin Heller's InfoWorld hands-on review of 6 November 2017 covers an early version. It provides historical context, not an assessment of version 2.5.2. Vendor funding was not disclosed in the report read.

Reported positives

  • The author described rapid automated model building.

Reported cautions

  • Data understanding and statistical knowledge remained necessary; historical prices are not current offers.

These reports reflect the linked authors’ experiences. Test date, plan and task affect results; they do not establish a representative overall rating.

Product development and notable changes

  • Version 2.5.2

    Release notes include experimental Linux ARM64 support and changes to Snowflake authentication. Check existing connectors before upgrading.

    Source for this change ↗

Which alternative fits which task?

DataRobot →

Compare for a broader platform including agent operations.

Akkio →

Consider for more guided marketing analytics.

Frequently asked questions

Is Driverless AI free and open source?

No. Its commercial license must be distinguished from other open H2O projects.

Does AutoML replace domain validation?

No. People must assess target definition, data quality, time-based separation and practical error costs.