Sourav Ganguly - ECE PhD Student of the Month - July 2025
Sourav Ganguly is a first-year Ph.D. student in the Helen and John C. Hartmann Department of Electrical and Computer Engineering, working in the Learning-Based Decision Making Lab as a research assistant under Dr. Arnob Ghosh. His research interests include machine learning, reinforcement learning, and stochastic control, with a focus on improving safety in autonomous systems operating in uncertain environments while maximizing long-term expected rewards. The goal is to support more effective control in real-world systems and reduce the gap between simulation and deployment.
Outside of academics, he enjoys reading, playing cricket and soccer, singing, cooking, acting, and mimicry. When not working on research, he’s usually reading a detective novel or watching a thriller series.
Autonomous systems are used in many domains, but most are trained in simulated environments. While simulated data includes sensor noise and other overhead, these systems often fail in real-world settings due to environmental variation and limited adaptability to uncertainty, leading to wasted resources. Sourav’s research addresses this by using the Robust Constrained Markov Decision Process (RCMDP) framework to improve control safety in uncertain Markovian environments.
What would you say that could be the next big thing in your area of research?
Reinforcement Learning (RL) in uncertain environments is set to advance significantly as real-world applications increasingly demand robustness to incomplete information, adversarial disturbances, and evolving dynamics. A central focus of future research will be distributionally robust RL, where agents learn policies that perform well under worst-case distributions within specified ambiguity sets—ensuring reliability in safety-critical areas like autonomous driving and health care. Another key direction is meta-RL and adaptive robustness, which aim to develop agents capable of generalizing across diverse uncertainty levels by learning robust behaviors from limited data. Emerging applications in finance, robotics, and energy will further fuel innovation, as robust RL techniques are needed to maintain performance amid market volatility, hardware malfunctions, or fluctuating demands. Lastly, the development of interpretable and verifiable robust RL methods will be crucial for ensuring transparency, regulatory compliance, and public trust. Ultimately, the future of RL in uncertain environments hinges on achieving a balance between robustness, adaptability, and efficiency to enable safe and dependable deployment in the real world.
You have taken courses from CS, DC and MATH department. Please share some experiences of taking non-ECE courses - how they have inspired your research, if they are harder or simpler, etc.
I had the opportunity to take advanced courses across a broad spectrum of subjects, which significantly strengthened the foundational knowledge essential for my research. Each course was intellectually rigorous, and I gained substantial insights from them. The mathematics courses, in particular, demanded more time and effort to master, especially since I do not come from a strictly mathematical background. These courses were especially challenging due to their emphasis on formal proof writing and deep theoretical reasoning. However, this challenge proved to be immensely rewarding in the long term. Compared to courses in Computer Science and Data Science, the Mathematics courses required greater time investment outside the classroom. Nevertheless, they greatly enhanced my confidence in constructing theoretical arguments—an essential skill for developing the theoretical framework and tools necessary for my research.
After spending the first year of your Ph.D. study at NJIT, what is one unique feature of your advisor that you have found to be most helpful?
My research advisor, Dr. Arnob Ghosh, is one of the most supportive individuals I have ever met. At times when I doubted myself especially during a demanding course load and research pressure he placed his trust in me and helped me discover a stronger, more resilient version of myself. His support extends beyond academic guidance; he has also been there for me during personal dilemmas, always offering thoughtful advice. Dr. Ghosh is exceptionally approachable and friendly, and every conversation with him leaves me with new insights. He is one of the most talented, knowledgeable, and disciplined people I’ve encountered. As I progress through my doctoral journey, I would consider myself fortunate to inculcate even a fraction of the skills and wisdom he has cultivated over the years. I am truly grateful to Dr. Ghosh and to NJIT for this opportunity, and I am committed to making the most of it.