Loss of life and property due to natural disasters (floods, wildfires, earthquakes, etc.) is an all too common occurrence each year. The speed and effectiveness of any disaster response operation relies heavily on two factors: 1) quality of situational awareness; and 2) time it takes to gather that situational awareness. Many traditional methods of monitoring disasters utilize manual inspections of satellite images, resulting in substantial delays and limited spatial coverage. In this paper, we introduce a real-time video intelligence framework that leverages deep learning object detection, semantic segmentation, and video stream analysis to provide actionable situational awareness during natural disaster events. .