Max Robotics — Global Service Robot Directory + US Certification

About

Lorenzo Pellegrini is a machine learning researcher specializing in continual learning, robotic vision, and efficient deep learning for resource-constrained systems. His work addresses one of the central challenges in modern AI: enabling neural networks to learn incrementally from new data without forgetting previously acquired knowledge — a problem known as catastrophic forgetting. Pellegrini's most influential contribution, "Latent Replay for Real-Time Continual Learning" (2020, 138 citations), pioneered practical strategies for training deep networks on embedded systems and robotic platforms, making continual learning viable in real-world edge computing scenarios. This work, alongside his research on rehearsal-free continual learning over non-I.I.D. batches (53 citations), demonstrates his focus on bridging theoretical advances with deployment constraints faced by physical robots operating in dynamic environments. His earlier work on fine-grained continual learning (2019) further established his presence in robotic object recognition, a domain he has helped shape through active community engagement. Pellegrini also co-organized the IROS 2019 Lifelong Robotic Vision Challenge, attracting over 150 competing teams globally, underscoring his role not only as a researcher but as a catalyst for collaborative progress in lifelong learning for autonomous systems.