H2O AI Cloud Demo
- Predictive AI
- H2O Driverless AI
Overview
This live demo walks through a full machine learning workflow inside H2O AI Cloud (now the H2O Sovereign AI Platform), using a Telco customer churn dataset. The presenter imports data through Drive, explores it with automatic visualizations, builds a supervised AutoML experiment in Driverless AI with interpretability prioritized for a customer-service use case, and reviews the model's Shapley-driven insights. The video closes by deploying the model through ML Ops as a live REST endpoint and wiring it into a custom H2O Wave app that gives a call center team real-time churn predictions.
Key takeaways
- How to bring data into H2O AI Cloud (now the H2O Sovereign AI Platform) once through Drive and reuse it across multiple engines (H2O-3, Driverless AI, Hydrogen Torch) without re-uploading.
- How to configure a Driverless AI experiment for a churn-prediction use case, prioritizing interpretability so a customer service team can see why a customer is flagged as a churn risk.
- How Driverless AI automatically checks for data leakage (for example, a "last payment date" style feature that would give away the answer) and runs threefold cross-validation to avoid overfitting on a smaller, 3,000-row dataset.
- How partial dependence plots surface actionable business rules, for example that customers who call support more than four times see a sharp jump in churn likelihood.
- How to register and deploy a trained model in ML Ops as a live REST endpoint, demonstrated here with a real-time curl request that returns a churn probability.
- How to build a no-code, Python-based front end with H2O Wave so a call center rep can look up a customer and get a live churn score without touching a model endpoint directly.
Start owning your AI