Case Study: Takenaka Corporation reduces worksite accidents with H2O.ai Driverless AI

Takenaka runs around 10 AI projects with H2O.ai to reduce worksite injuries

Takenaka Corporation, one of Japan's largest general contractors, faced the challenge of building and deploying numerous AI models with a very small analytics team. Their goal was to predict occupational accidents on worksites to improve safety, but outsourcing was costly and an earlier AutoML solution hindered adoption. They turned to vendor H2O.ai and its product H2O Driverless AI to automate model development and enable flexible deployment across different environments.

The solution from H2O.ai used H2O Driverless AI to automate feature engineering and build a model that predicts the types of accidents most likely to occur based on worksite data. This model is deployed via REST API, MOJOs on containers, or embedded directly in systems to provide predictions both online and at offline worksites. As a result, Takenaka's small team now runs approximately 10 AI projects in production. The accident prediction model has successfully reduced worksite injuries and fatalities, and its accuracy matches or beats models built manually by data scientists.

Takenaka Corporation