Max Robotics — Global Service Robot Directory + US Certification

About

Murtaza Dalal is a robotics and machine learning researcher whose work sits at the intersection of reinforcement learning, robot manipulation, and self-supervised learning. His research focuses on enabling autonomous agents to acquire broad, generalizable skills with minimal human supervision — a fundamental challenge in building practical robotic systems. Dalal's most influential contribution, "Visual Reinforcement Learning with Imagined Goals" (2018, 183 citations), demonstrated how agents could learn goal-conditioned behaviors directly from raw image observations, significantly advancing goal-directed autonomy. This line of work continued with "Skew-Fit" (2019, 66 citations), which tackled self-supervised goal setting to encourage more thorough environment exploration. His work on "AWAC" (2020, 71 citations) addressed a critical bottleneck in real-world robotics by combining offline datasets with online fine-tuning to dramatically accelerate learning. Beyond foundational RL, Dalal has explored action primitive representations, sim-to-real transfer through local manipulation policies, and language model-guided planning for long-horizon tasks. His most recent work, "Plan-Seq-Learn" (2024) and "ManipGen" (2025), reflects a growing interest in integrating large language models with RL to tackle complex, multi-step robotic challenges. Collectively, his research has meaningfully shaped how robots learn to perceive, plan, and act in unstructured environments.