The increasing availability of mobile phones with embedded GPS devices and sensors has spurred the use of vehicle telematics in recent years. Telematics provides detailed and continuous information of a vehicle such as the location, speed, and movement. Vehicle telematics can be further linked with other spatial data to provide context to understand driving behaviors. The collection of high-frequency telematics data results in huge volumes of data that must be processed efficiently. We present a solution that uses Apache Spark to load and transform large-scaled telematics data. We then present how to use machine learning on telematics data to derive insights about driving safety.