Hands-on experience advancing egocentric video understanding with large-scale first-person datasets, focusing on novel representations, head-motion robustness, long-term prediction, memory querying, and cross-domain transfer for AR, robotics, and assistive systems.