Make with H2O.ai: Getting Started with H2O Driverless AI
- Predictive AI
- H2O Driverless AI
Overview
An H2O.ai data scientist walks through Driverless AI, the company's automated machine learning platform, and previews a new Predict Wizard feature built on the open-source Nitro app framework. The session traces how H2O's tooling evolved from open-source H2O-3 and AutoML to Driverless AI's Python-based automation, then runs a live comparison between the standard one-click Predict flow and the new guided wizard on an airline-sentiment dataset. The wizard interactively flags data leakage that a plain AutoML run would otherwise bury in a warning.
Key takeaways
- How Driverless AI evolved from H2O's open-source Java platform and H2O AutoML into a Python-based tool that orchestrates XGBoost, TensorFlow, and other libraries
- Why a single-button AutoML run can produce a technically accurate but useless model, shown live with a dataset column that leaks the answer
- How the new Predict Wizard, built on the open-source Nitro app framework, interactively surfaces data leakage, asks about production constraints like Java deployment, and checks whether GPU-based BERT models are feasible
- How to compare the standard Predict flow against the Wizard flow side by side on the same airline-sentiment dataset
- What's coming next for Driverless AI, including dataset joins and a business-value calculator for evaluating model profit
Start owning your AI