Ricerc@Sapienza

My research integrates generative AI, 3D computer vision, and robotics to develop photorealistic and physically accurate simulators reconstructed directly from in-the-wild videos. The primary goal is to close the sim-to-real loop, enabling robust policy learning, reproducible benchmarking, and natural language debugging of real-world robotic systems. Ultimately, I aim to push this concept to its extreme by reconstructing entire cities from satellite videos, thereby extending embodied simulation and control to truly global scales.