Embodied Neuromechanics or: How I Learned to Stop Worrying and Love NeuroAI

Behavior is the ultimate output of the nervous system, but it has long resisted precise measurement. My lab develops AI tools, including SLEAP, that extract detailed body kinematics from video at scale, turning behavior into quantitative data on par with neural recordings and connectomes. We then use these measurements to build embodied neuromechanical models: artificial neural networks that control physically simulated animal bodies, linking neural circuits, biomechanics, and behavior in a single framework. Because these models are generative, they can simulate perturbations and produce mechanistic hypotheses that can be tested directly at the bench, creating a feedback loop between AI and experimental neuroscience.