My research is driven by the goal of developing autonomous embodied agents which can intelligently interact and learn in the physical world. My research has mainly focused on implicit learning through energy-based models. My work has focused on how such an implicit learning learning procedure enables compositional generative models which can both compositionally generate novel outputs and compositionally understand the world. My work has further focused on how such compositional models may be applied to trajectory planning, enabling flexible adaptation to novel goals and rewards across both synthesized videos and on the real robotic settings.