Max Robotics — Global Service Robot Directory + US Certification

About

Pietro Mazzaglia is a robotics and machine learning researcher whose work sits at the intersection of reinforcement learning, world models, and embodied AI. His research addresses fundamental challenges in enabling robots to perceive, reason about, and interact with their environments more efficiently and intelligently. Mazzaglia has made notable contributions to object-centric representation learning, developing approaches that allow robotic agents to parse complex scenes into structured, manipulable object representations rather than relying on monolithic global state encodings. His FOCUS framework exemplifies this direction, proposing world models that explicitly capture object-level structure to improve robotic manipulation — work that has already attracted early citation interest. His research on Active Inference-based scene understanding further demonstrates his commitment to biologically inspired, uncertainty-aware perception. Beyond representation learning, Mazzaglia has tackled practical deployment challenges, including computational optimization of image-based reinforcement learning for resource-constrained robotic hardware, and investigating how action space design — particularly redundancy-aware formulations — affects learning efficiency in robot arms. With a growing body of work accumulating citations across multiple venues, Mazzaglia represents an emerging voice in robot learning, bridging cognitive science-inspired modeling with real-world robotic applicability. His portfolio makes him a compelling researcher to follow for students interested in next-generation autonomous robotic systems.