Unmanned aerial vehicles (drones) have the potential to transformatively impact numerous industries, including infrastructure inspection, search and rescue, package delivery, and even atmospheric measurements. To do so, drones must be able to perform reliably in real-world weather conditions, including extreme wind and gusts. Modern drones do not actively sense the evolving wind conditions; instead, they react to the wind as it perturbs them. To address this, we design and implement MAST (MEMS Anemometry Sensing Tower), a fast, accurate, omnidirectional flow sensor. We demonstrate that MAST improves drone performance in windy conditions: improving the drone’s reaction time and reducing error. To do so, we train a ‘wind-aware’ controller using reinforcement learning in simulated wind. In extensive hardware experiments, we demonstrate the wind-aware controller outperforming two strong ‘wind-unaware’ baseline controllers in challenging windy conditions.