Voron

Artificial Intelligence in Aerospace is a new student organization at Georgia Tech, with 25+ engineers and one flagship program, the Voron drone project. The premise is to own the whole autonomy stack, from the airframe to the policy that flies it. I'm on the hardware team and I own the mesh network.

Voron is a loop. Each physical airframe has a digital twin matched to its mass, motors and sensors. Policies train against the twin by reinforcement learning, deploy to the real drone, and every flight feeds telemetry back into the twin. Perception runs on the fleet's own EO/IR footage. The vehicles talk over a self-healing wireless mesh instead of a single ground link: every drone is a relay, and the topology re-forms when a node drops out. That radio layer is mine.

The simulation side already runs: a quadrotor physics model, a photoreal Unreal Engine city, and a PPO pipeline graded on fixed-seed episodes with noisy sensors and injected faults. The airframe above is the same model in the simulator, the renderer, and on the bench. My job is the last one: building the vehicle that matches its twin, and the network that keeps the fleet talking.