The Fire-observation aerial robots of RAINDROPS integrate a multi-modal sensing payload tasked with the goal of detecting and localizing incipient-stage wildfires from high-altitude. The system relies on deep learning applied across sensor signatures and a custom actuated perception solution to increase area coverage. The fire-observation aerial robots of RAINDROPS shall operate in a team of systems with multiple such units operating simultaneously in order to cover a 1,000km2 area rapidly.