Wildfires have been a significant concern for communities and fire response agencies in many countries. Hence, it is critical to be able to predict the fire risk in a timely and accurate manner and at granular level. However, this requires accessing and processing large amounts of spatial and temporal data from a number of sources in near real-time, while ensuring the immediate availability of risk measurement results. In this paper, we describe a large-scale data-driven system for personalized risk mitigation, fire response's resource optimization and dynamic evacuation planning. It leverages large spatial and temporal datasets to provide predictive analytics in near real-time and to deliver tailored insights to government agencies, communities and individuals.