My research interests lie in developing machine learning tools to create autonomous embodied agents. Two major challenges presented in the embodied setting include: high uncertainty and the rich combinatorial nature of the world. To tackle these issues, I utilize energy-based models to model underlying probability distributions in this setting and to construct composable models that can quickly adapt and generalize to new experience. Moreover, I use the energy optimization procedure as an adjustable computational budget, allowing longer computation times to adapt to novel out-of-distribution experiences. Embodied learning is multimodal, and I use neural fields as a tool to capture structure across modalities such as vision, text, sound, and touch. Finally I'm interested in broader applications of these tools to other domains such as computational biology.