Max Robotics — Global Service Robot Directory + US Certification

About

Rugile Pevceviciute is a roboticist whose research sits at the intersection of meta-reinforcement learning, self-supervised perception, and generalist robotic manipulation. Her work addresses two fundamental challenges in robotics: enabling rapid adaptation to new tasks and building scalable, multi-embodiment learning systems. In her highly cited paper "Offline Meta-Reinforcement Learning for Industrial Insertion" (66 citations), she pioneered methods that allow robots to adapt to novel insertion tasks with minimal trials by leveraging prior task experience—a critical step toward practical industrial deployment. Pevceviciute also introduced S3K (Self-Supervised Semantic Keypoints), a framework that learns consistent visual keypoints across multiple camera views without human labels, enabling robust perception for manipulation. Most notably, she is a core contributor to RoboCat, a self-improving generalist agent that learns across diverse robots and tasks, representing a significant leap toward foundation models for robotics. Her latest work, DemoStart, combines demonstration-guided curricula with sim-to-real transfer for dexterous multi-fingered hands, achieving complex behaviors from sparse rewards. With over 90 total citations and publications spanning top venues, Pevceviciute is shaping the future of adaptable, data-efficient robot learning.