Eyeware creates natural and intuitive human-machine interaction experiences using eye-gaze tracking. We propose to design a novel, robust gaze estimation system based on multi-task deep learning that can adapt to challenging conditions, targeting new applications in home automation and advertising.