We demonstrate a robot-assisted feeding system that enables people with mobility impairments to feed themselves. Our system design embodies Safety, Portability, and User Control, with comprehensive full-stack safety checks, the ability to be mounted on and powered by any powered wheelchair, and a custom web-app allowing care-recipients to leverage their own assistive devices for robot control. For bite acquisition, we leverage multi-modal online learning to tractably adapt to unseen food types. For bite transfer, we leverage real-time mouth perception and interaction-aware control. Co-designed with community researchers, our system has been validated through multiple end-user studies.