Despite decades of progress in robotic hardware, sensing, and motion control, the autonomous deployment and scalable operation of robots in real-world field environments remain limited. Persistent challenges arise from uncertainty, risk, and the physical constraints of dynamic, unstructured settings. In this talk, Dr. Ali Agha, CEO of FieldAI, will describe how a physics-first foundation model designed specifically for Physical AI can address long-standing autonomy challenges in robotics. Unlike conventional vision or language models adapted for robotic systems, FieldAI’s Field Foundation Models are developed from the ground up to reason about the physical world and operate under real-world constraints. This approach enables robots to make decisions in real time and navigate dynamic, unstructured environments without relying on maps, GPS, or predefined paths. Dr. Agha will outline how this architecture-first approach to foundation model development is already being deployed and scaled successfully in industrial environments around the world.