My research vision is to enable collaborative autonomous intelligence by empowering robots with human-level spatial, social, and self-awareness, allowing them to actively perceive and plan in unstructured environments, interact effectively with humans or other robots, and leverage as well as augment the associated memory. To this end, I draw from vision, learning, robotics, graphics, language, sensing, data science, and cognitive science. My research works include developing robust, efficient, and scalable computational models for 3D scene parsing and decision-making from high-dimensional sensory input, as well as curating large-scale datasets to effectively train and verify these models for practical applications, including but not limited to connected and autonomous vehicles, assistive and service robotics, and construction automation.