Max Robotics — Global Service Robot Directory + US Certification

About

Razvan Pascanu is a prominent machine learning researcher whose work spans continual learning, physics simulation, and robot learning. He is perhaps best known for his contributions to continual learning in deep neural networks, exploring how artificial systems can learn incrementally from sequential experience in the way humans do — a challenge that remains central to modern AI research. His 2020 review on continual learning has garnered over 450 citations, reflecting its significance as a foundational reference in the field. Pascanu has also made notable strides in physics-based reasoning, co-developing Visual Interaction Networks — models that enable agents to predict the future states of physical systems directly from video, accumulating nearly 260 citations across related publications. His work on sim-to-real transfer in robotics, particularly using progressive networks to bridge the gap between simulated training and real-world deployment, has further cemented his influence across reinforcement learning and robotics communities. More recently, Pascanu has contributed to large-scale action models for robotics, exploring alternatives to Transformer architectures for efficient inference. His benchmarking efforts, such as Continual World, underscore his commitment to rigorous evaluation of continual reinforcement learning agents, making him a vital voice in shaping the field's empirical foundations.