Max Robotics — Global Service Robot Directory + US Certification

About

Jodilyn Peralta is a leading researcher at the intersection of robotics, machine learning, and real-world control, whose work is defining how robots learn to operate in unstructured environments. Her primary contributions center on scaling robot learning through innovative architectures and data augmentation techniques. She is the lead author of the highly influential "RT-1: Robotics Transformer for Real-World Control at Scale," which has garnered over 512 citations. This seminal work demonstrates how large, diverse, task-agnostic datasets can be leveraged to enable robots to solve specific tasks with minimal fine-tuning, a paradigm shift from traditional, task-specific training. Peralta further advanced the field with "Scaling Robot Learning with Semantically Imagined Experience" (66 citations), a novel approach that uses semantic imagination to generate synthetic training data, dramatically increasing the scale and diversity of robot learning datasets without requiring additional physical robot time. Her work on "Q-Transformer" also provides a scalable method for offline reinforcement learning, enabling multi-task policies to be trained from both human demonstrations and autonomously collected data. Through these contributions, Peralta is a key architect of the modern, data-driven approach to generalist robotics.