Hi, I am Abdus Salam Azad, a final-year Ph.D. student at UC Berkeley (expected graduation August 2024). I have been incredibly fortunate to be advised by Prof. Ion Stoica. I also have the privilege of collaborating with Prof. Pieter Abbeel from UC Berkeley and Izzeddin Gür & Aleksandra Faust from Google DeepMind. My research has been funded by Google. Inc. & BAIR Commons Research initiative. 

I am passionate about designing intelligent autonomous agents to solve complex and realistic tasks that require sequential decision making. During my Ph.D., I focused on Environment Generation & Curriculum Learning methods that algorithmically generate diverse training environments (e.g., learning scenarios) to improve the generalization of agents trained with Reinforcement Learning while improving sample efficiency. Recently, I have been focusing on designing 'agentic' workflows to solve complex sequential decision making tasks (e.g. web navigation) with multi-modal Large Language Models. In this era of rapidly advancing multi-modal LLM agents and their increasing application demand, it is crucial to develop principled methods and frameworks to ensure these agents are safe, responsible, and high-performing.
Previously, I completed my Bachelor of Science in July 2014 and Masters of Science in January 2017 from CSE, BUET. During my M.Sc., I worked on improving the exploration-exploitation tradeoff of Memetic Algorithms—a variant of Genetic Algorithms—to solve complex real-world Combinatorial Optimization problems.

Please find my resume here.

Thanks for visiting my site. Have a good day. 

Selected Publications (C: conference / J: journal / P: preprint *: equal contributions)

[C] CLUTR: Curriculum Learning via Unsupervised Task Representation Learning [arxiv] [code]

[C] Programmatic Modeling and Generation of Real-time Strategic Soccer Environments for Reinforcement Learning [preprint] [code]

[J] A Heuristic Initialized Stochastic Memetic Algorithm for MDPVRP With Interdependent Depot Operations [pdf] [code]