I’m Irvin Dalaud, a Master’s student in Robotics at EPFL (École Polytechnique Fédérale de Lausanne), with a deep-rooted passion for neuroengineering and human-centered robotics. Born and raised by an amputee mother, I’ve long been driven by a personal mission: to bring advanced assistive technologies—such as robotic prosthetics and exoskeletons—from the lab to the real world.
My work bridges robotics and neuroengineering, with experience in EMG-based force control of dexterous hands (supervised by Prof. Josie Hughes and Prof. Silvestro Micera), and actuated knee joint design in collaboration with Prof. Abdul Barakat.
Alongside research, I’m a Venture Capital Analyst at Elaia, focusing on deep tech startups, and a Research Assistant at HEC Lausanne, working on large-scale data analysis.
MSc Mechanical Engineering
École Polytechnique Fédérale de Lausanne
International Exchange
École Polytechnique, l'X
BSc Mechanical Engineering
École Polytechnique Fédérale de Lausanne
I’m interested in building intelligent robotic systems by combining model-based control, reinforcement learning, and biosignal-driven interfaces. I enjoy working on problems that involve high-dimensional continuous control, where techniques like MPC and deep RL can be used alongside EMG signals to enable natural and adaptive human-robot interaction. I’m particularly drawn to applications in dexterous manipulation, locomotion, and prosthetic control.
Always happy to connect and discuss ideas!
Implementation of an Neural Network based EMG decoder for hand prosthesis, robust to time-varying muscle signal properties due to muscle fatigue. Python code based on Gaussian-Mixture Models and Biomedical signal processing and filtering techniques.
Implementation of locomotion controllers based on Central Pattern Generators (CPG) or Deep Reinforcement Learning for a quadruped robot. RL policy is developed using MuJoCo for simulation and PPO for training.
Design of a robotic knee joint replicating key biomechanical functions of the biological knee. The powered knee is based on a Virtual Rolling Sphere contact mechanism, and through an optimization routine coupled with a parametric CAD is able to adapt to any size requirements.