I've led SafeShot at Mila's AI4Good Lab, reproducing EditShield in PyTorch and building a local image-protection tool that blocked 73 of 85 manually verified AI editing threats while preserving facial identity.
Across ROS2 robotics, STM32 firmware, and full-stack AI prototypes, I build complete systems: an LLM-to-action pipeline for a Kinova arm, autonomous drone security architecture, AR cognitive assistance, and real-time embedded sensor applications.

