Post-doctoral research associate at the University of Glasgow on the DIFAI project.
Developed an Active Inference model for simulating human-like mouse point-and-click behavior by minimizing Expected Free Energy.
Co-developed SIM2VR, a system enabling closed-loop alignment between biomechanical user simulation and VR applications, allowing simulated users to be trained directly in real VR environments.
Built SimMPC, a Model Predictive Control framework for simulating mid-air pointing interactions using a second-order muscle model, achieving high prediction accuracy for individual user movements.
Proposed the 'User in the Box' (UitB) approach, using deep reinforcement learning to train perceptually controlled biomechanical models on four increasingly complex HCI tasks.
Applied Optimal Feedback Control theory (OFC4HCI) to unify human body and computer dynamics in pointing tasks, releasing a Python toolbox for parameter identification.
Explored reinforcement learning control of a full upper-extremity biomechanical model in MuJoCo, incorporating signal-dependent and constant motor noise to predict 3D reaching movements.