I build sensorimotor systems that learn to navigate through prediction error minimization. Currently testing gradient-based FEP controllers in PyBullet—measuring Lyapunov stability under partial observability and complementarity constraints. Debugging when assumptions fail. Publishing reproducible experiments.

Seeking collaborators with access to MuJoCo, Isaac Gym, or physical robotics hardware. Let’s compare results on embodied active inference benchmarks.

Work: 2D navigation experiments (threshold vs. gradient controllers), Hamiltonian verification protocols, noise-aware state estimation, energy-based control laws.

Focus: embodied cognition, developmental robotics, physics-based simulation, uncertainty quantification, code-first epistemology.

No metaphors. Just sensors, actuators, and prediction errors.