Max Robotics — Global Service Robot Directory + US Certification

About

Jack Parker-Holder is a rising leader at the intersection of reinforcement learning (RL), neural architecture search (NAS), and robust optimization. His research focuses on making RL more scalable, sample-efficient, and generalizable—particularly for real-world robotics and vision-based decision-making. Parker-Holder’s most influential contribution is the development of **ES-ENAS**, a method that elegantly combines Evolution Strategies (ES) with Efficient NAS to automatically discover compact, high-performing RL policies without additional computational cost. This work, along with his earlier paper on provably robust blackbox optimization, has helped bridge the gap between derivative-free methods and modern deep RL. He has also advanced **zero-shot dynamics generalization** from offline data, enabling agents trained in a single environment to adapt to unseen conditions—a critical step toward deploying RL in safety-critical domains. More recently, his work on **implicit attention** for pixel-based RL and the provocative concept of “video as the new language” for decision-making signals a forward-looking agenda. With over 30 citations across his top papers and growing recognition, Parker-Holder is shaping how we design, train, and trust RL agents in complex, high-dimensional worlds.