Anh-Quan Pham

Anh-Quan (Bill) Pham

You can just call me Bill or Quan.

I am an AI researcher at Dyna Robotics, where I work on commercial-grade robotics foundation models. I enjoy thinking about how scalable robot learning, reinforcement learning, and foundation models can create reliable, general-purpose manipulation robots in the real world.

I did my graduate degree in robotics at the GRASP Lab, University of Pennsylvania. There, I was fortunate to be advised by Professors Dinesh Jayaraman, Eric Eaton, and Kostas Daniilidis, and to collaborate closely with Professors Jorge Méndez‑Méndez, Dani S. Bassett, and Osbert Bastani. I am incredibly grateful to have been mentored by Marcel Hussing in reinforcement learning, and by Junyao Shi and Subin Kim in robot learning.

During my time at Penn, I served as President of the Penn Robotics Entrepreneurs Club (PREC), connecting aspiring founders and researchers with industry leaders and top investors across the US, Europe, and Asia, turning bold ideas into real ventures.

Before Penn, I was a researcher at the Institute for Infocomm Research, A*STAR Singapore, studying interpretable reinforcement learning under Dr. Senthilnath Jayavelu. During my 2.5 years of undergrad at VinUniversity, I applied RL to next-generation telecommunication networks under Professor Van‑Dinh Nguyen, exploring resource allocation and network slicing.

Selected Work

§ Equal contribution (unless noted). † Equal advising. ‡ Research project lead.

Iterative Compositional Data Generation for Robot Control

Anh-Quan Pham, Marcel Hussing, Shubhankar P. Patankar, Dani S. Bassett, Jorge Mendez‑Mendez, Eric Eaton

Transactions on Machine Learning Research (TMLR), 2026

Summary Website Journal/Publisher arXiv PDF Code OpenReview X Bluesky

ASSET2SIM: Automatically Improving Simulation Physics for Articulated Objects

Anh-Quan Pham^§, Sagnik Anupam^§, Luyang Hu^§, Kaitian Chao, George Gao, Tianyou Wang, Junyao Shi, Osbert Bastani^†, Dinesh Jayaraman^†

§ Sorted in alphabetical order by last name; modified for clarity.

Under Review

Summary

Zero-Shot VLA (Vision-Language-Action model) Adaptation with Retrieved Guidance

Subin Kim, Anh-Quan Pham, Dinesh Jayaraman

Under Review

Summary

SBAMP: Sampling Based Adaptive Motion Planning

Shreyas Raorane^§, Kabir Ram Puri^§, Anh-Quan Pham^‡

Workshop on Frontiers of Optimization for Robotics, International Conference on Robotics and Automation (ICRA), 2026

Summary Journal/Publisher arXiv PDF Code OpenReview Video

News & Updates

April 2026: 🚀 I've decided to join the AI Research team at DYNA Robotics in Silicon Valley this June to build commercial-grade physical AGI. 🦾 🧠

April 2026: 🎉 Excited to share that SBAMP: Sampling-Based Adaptive Motion Planning has been accepted to the Workshop on Frontiers of Optimization for Robotics (2nd Edition) at ICRA 2026! 🤖 This marks my first project in a research leadership role (last author). Grateful for the team that made this possible. 🦾

April 2026: 🎉 Exciting news: Iterative Compositional Data Generation for Robot Control has been accepted to Transactions on Machine Learning Research (TMLR)! Really grateful to the TMLR reviewers and action editor for their sharp, constructive feedback. 🙌 Check out our project website and the open discussion on OpenReview.

December 2025: Our new paper, Iterative Compositional Data Generation for Robot Control, is out! We are excited to open-source the code and share the work. Try it out on GitHub and join the discussion on Twitter/X and Bluesky.

May 2025: I will serve as President of the Penn Robotics Entrepreneurs Club (PREC) during the 2025–2026 year!

January 2025: Excited to share that I've started as a Graduate Research Assistant at the GRASP Lab, University of Pennsylvania, working on various aspects of Reinforcement Learning for Robotics!