Drone imagery is already widely used in military operations, but the bottleneck is analysis: raw video requires human interpretation, which is slow and cognitively demanding. Running AI classification models on the drone itself — at the edge — allows processed intelligence (vehicle type, personnel count, anomaly flags) to reach the company commander without relying on a bandwidth-intensive video feed. The project asks: what target classification tasks can be performed reliably by an edge AI model on a small drone platform, and how should the processed output be formatted for a company commander who needs a summary, not raw footage? The team would work with open aerial imagery datasets, train and compress a detection model suitable for edge deployment, and prototype the commander's intelligence summary interface. The deliverable is a prototype edge AI classification pipeline with a documented accuracy-versus-model-size trade-off analysis and a commander-facing summary display.