Throughout my career I've always been interested in the most challenging topics in computer vision and machine learning: 3D reconstruction, video registration, egocentric vision, self-supervised optical flow, generative adversarial networks or neural architecture search to name a few.

At Panasonic β AI Laboratory, Panasonic top deep learning research group in the US, I developed a novel raindrop removal method that leverages the power of attention and generative adversarial networks to clean rainy video sequences, enabling outdoor computer vision for robotics, autonomous driving, surveillance etc. My work has been published at one of the top venues for autonomous driving and, far from being just a research prototype, has been used to deliver real-world demos in multiple international settings.

My constant strive for improvement and challenge led me to a visiting researcher position at MILA - Quebec Artificial Intelligence Institute. In this unique blend of industry and academia I am working with world renowned students, post-docs and professors on one of the hottest topics of machine learning: neural architecture search. My current goal is to develop a method that automatically designs and optimizes neural networks for multimodal data, relieving engineers from the burdensome task of manual architecture engineering and enabling the next generation of multimodal learning algorithms.