I work at FAIR at Meta.

I can be contacted by yangtsungyen [at] gmail.com / Google Scholar / LinkedIn / Twitter

I got my Ph.D. in the Department of Electrical and Computer Engineering at Princeton University, working with Prof. Peter Ramadge and Prof. Karthik Narasimhan between 2017 and 2022.

My research lies at the intersection of machine learning (ML), reinforcement learning (RL), and large foundation models such as large language models (LLM), and vision language action models (VLA). My goal is to develop state-of-the-art artificial intelligence (AI) systems that enable safety assurance, scalability, and personalization. Some applications of my research include (1) robotics: home robots, dexterous hand manipulation, self-driving cars, robot locomotion, human-robot interaction; (2) LLM AI agent: web agents, instruction following, language grounding; and (3) time-series data analysis: AI-based traffic control, anomaly detection, domain adaptation, and personalized learning in education. My work has been published in leading AI venues, including NeurIPS, ICML, ICLR, EMNLP, ECCV, CoRL, IROS, etc.

At Meta, I have been working on training and deploying Vision Language Action Models (VLA) and Large Language Models (LLM) AI Agent. Specifically, I 

• Lead the team and develop state-of-the-art VLA models for home robots and dexterous hand manipulation using imitation learning and reinforcement learning.

• Design an embodied AI benchmark -- Partnr and Habitat 3.0 for planning and reasoning in embodied multi-agent tasks with high-quality simulation realism. 

• Build a hierarchical AI stack for robot -- Adaptive Skill Coordination (ASC) for accomplishing long-horizon tasks such as mobile pick-and-place.

• Develop a framework -- HomeRobot: an affordable compliant robot that navigates homes and manipulates a wide range of objects in order to complete everyday tasks.

• Research on AI agents that solve programming issues using VSCode interface via self-learning approaches.

• Have two media launches:  (1) Advancing machine intelligence through human-centered research, AI at Meta channels on X, Facebook, and YouTube. It was also announced by Mark Zuckerberg, Feb 7, 2025, and (2) Advancing embodied AI through progress in touch perception, dexterity, and human-robot interaction, AI at Meta blog. It was also announced by Mark Zuckerberg and Chris Cox, Oct 31, 2024.

Before joining FAIR, I have been working on several ML/RL/NLP/time series data analysis projects , including but not limited to

(1) Reinforcement learning with safety constraints for self-driving cars;

(2) Learning control policy for robotics systems;

(3) Anomaly detection from sensor measurements;

(4) Instruction following in natural language processing;

(5) Simulation to real transfer using physics priors and transformers;

(6) Representation learning from user behavioral data.

Before Princeton, I received my BS in Electrical Engineering and Computer Science (Honors Program) at National Chiao Tung University, Taiwan.