Anh-Quan Pham

Welcome to my website. Please feel free to read more about me, or you can check out my resume, projects, view site statistics, or contact me.

I am a second-year master's student in Robotics at the University of Pennsylvania, where my research centers on reinforcement learning (RL) as the foundation of Physical AI. I use RL to study how agents explore, adapt, and continuously improve their actions in the real world. My goal is to move beyond reward maximization and benchmarks, and build agents that learn and explore efficiently in the wild, generalize across tasks, and act in ways people can understand and trust.

At Penn, I work on several directions that reflect this vision:

• Under the guidance of Professors Eric Eaton, Jorge Méndez Méndez, and Dani S. Bassett, I study compositional zero-shot data generation, where agents can tackle new tasks by recombining prior knowledge.
• Supervised by Professors Dinesh Jayaraman and Osbert Bastani, I develop LLM-guided reward design, conduct large-scale RL training for dexterous tool use with the Franka arm, and work on articulated simulation alignment with real-world physics.
• I also learn a lot from the mentorship of Marcel Hussing, who focuses on stable and reliable RL, and Junyao Shi, who brings internet-scale data and foundation models into robotics. Their guidance reflects my vision of reliable and scalable Reinforcement Learning as the driver of continuous improvement and sim-to-real transfer, supported by developments in foundation models to leverage internet data for high-level world understanding.

Before Penn, I studied interpretable reinforcement learning at A*STAR Singapore under Dr. Senthilnath Jayavelu, focusing on symbolic policies and latent representations to improve transparency in decision-making. As an undergraduate at VinUniversity with Professor Van-Dinh Nguyen, I applied RL to next-generation telecommunication networks, exploring resource allocation and network slicing, and completed my thesis on adaptive robotic parameter optimization using RL. As an engineering lead intern at Huawei Vietnam, I applied machine learning to IoT systems, leading a project on sleep-stage classification and representing Vietnam at the Asia-Pacific Seeds for the Future Summit.