Masudur Rahman

I specialize in Reinforcement Learning and Robotics, focusing on creating intelligent agents that can make critical decisions under uncertain conditions. My current work centers on generalization in reinforcement learning, with the aim of developing algorithms that are robust against confounders in real-world tasks. I am investigating the use of foundation models in decision-making, exploring their potential in applications such as surgical robotics, medical imaging, and the automation of burn injury treatment.

I am currently a Postdoctoral Research Assistant in the Edwardson School of Industrial Engineering at Purdue University, working with Dr. Juan P. Wachs. I completed my Ph.D. in Computer Science at Purdue University in 2024 under the supervision of Dr. Yexiang Xue. I completed my M.S. in Computer Science at the University of Virginia in 2018. Before that, I worked as a Lecturer at BRAC University from 2013 to 2015, after earning my B.Sc. in Computer Science and Engineering from BUET in 2013.

KEY RESEARCH AREAS
Reinforcement Learning: I develop reinforcement learning (RL) algorithms to cater to the needs of surgical decision-making and robotic surgery. My methods include enhancing generalization through style transfer, increasing sample efficiency via novel policy gradient techniques, and using natural language for interpretable policy training. These approaches address challenges such as distribution shifts and primacy bias, demonstrating superior generalization and performance in various testing environments.
AI in Burn Care: I have developed an AI system based on the Vision-Language Model (GPT4-Vision) for automated surgical decision-making in burn patients. This system enhances preoperative decisions, ensuring timely and accurate surgical interventions, and has demonstrated performance levels surpassing expert surgeon capabilities.
Teleoperated Robotic Surgery: My work includes developing semi-autonomous telesurgery systems that can function effectively despite communication delays (5s compared to 300ms for direct teleoperation) and unreliable connections. These systems make remote surgical operations feasible, providing critical surgical care to remote, austere and underserved areas.

FUTURE GOALS
Expanding Access to Surgery: In the near term, my goal is to develop systems that allow general medical personnel, such as nurses and medics, to perform surgical procedures with AI and robotic assistance. This approach aims to reduce the burden on hospital specialists, increase access to surgical care, and lower healthcare costs.
Autonomous Surgical Systems: Looking further ahead, I aim to create fully autonomous surgical systems capable of managing entire surgical processes. A prototype in development, RoBurn—an Automated Robotic Burn Surgeon—is equipped with technologies like ultrasound and digital cameras to perform comprehensive burn care, even in challenging environments.

COLLABORATION
I am fortunate to collaborate with a diverse group of disciplines, including engineering, robotics, hospitals and medical facilities (IU, UPMC), surgeons, and military personnel through various research projects. I am the student team lead for the AMBUSH project, an interdisciplinary collaboration involving the Department of Computer Science and School of Industrial Engineering at Purdue University, as well as the Department of Surgery at the School of Medicine, University of Pittsburgh (UPMC). We are developing an AI system for burn care with active collaboration from Gayle Gordillo, Professor of Plastic Surgery and Director of Wound Care, and Mohamed Salah El Masry, Assistant Professor of Surgery. My Ph.D. research is supported by grants from the NSF, NIH, and the Department of Defense (DoD).