cv

Master of Science

Stanford University

M.S. Computer Science

• Specialized in Artificial Intelligence

Bachelor of Science

University of California, San Diego

B.S. Computer Science | Regents Scholar | Summa Cum Laude

• [Selected Courses] CS: Data Structures, Algorithms, Software Engineering, Database, Operating System, Computer Security, Network Services; 3D/CG: Computer Graphics, 3D User Interaction, 3D Asset Design; AI: Statistical Methods, Probabilistic Models, Machine Learning, Deep Learning, Computer Vision I/II, ML for Music/Audio, ML with Few Labels, Deep Learning for 3D Data (Graduate), ML for Robotics (Graduate)
• [Activities] Director of Events at ACM AI, ICPC Team, Scholars Society, Intern at SU Lab, Intern at Qualcomm Institute

Demonstrating GPU Parallelized Robot Simulation and Rendering for Generalizable Embodied AI with ManiSkill3

Stone Tao, Fanbo Xiang, Arth Shukla, Yuzhe Qin, Xander Hinrichsen, Xiaodi Yuan, Chen Bao, Xinsong Lin, Yulin Liu, Tse-Kai Chan, Yuan Gao, Xuanlin Li, Tongzhou Mu, Nan Xiao, Arnav Gurha, Viswesh N., Yong Woo Choi, Yen-Ru Chen, Zhiao Huang, Roberto Calandra, Rui Chen, Shan Luo, Hao Su

Robot Learning Workshop at ICLR 2025 (Oral). Robotics: Science and Systems (RSS) 2025

RFCL: Reverse Forward Curriculum Learning for Extreme Sample and Demonstration Efficiency in RL

Stone Tao, Arth Shukla, Tse-kai Chan, Hao Su

International Conference on Learning Representations (ICLR) 2024

AI Research Intern

Advisor: Prof. Hao Su

Python, PyTorch, JAX, Gymnasium, Docker, Kubernetes

• Researched demo-guided deep reinforcement learning methods to effectively solve long-horizon, sparse tasks. Co-authored paper Reverse Forward Curriculum Learning accepted in ICLR 2024.
• Benchmarked various state-of-the-art demonstration-guided deep RL methods, including RLPD, IQL, etc., on ManiSkill2, D4RL, and Meta-World tasks. Performed experiments on Kubernetes cluster using Docker.
• Adapted TD-MPC2 to Maniskill3 CPU/GPU vectorized environments and visual (rgb) based RL.
• Developed and optimized implementations of TD-MPC2 and SAC in JAX, achieving a 5x reduction in training time in comparison to previous PyTorch implementations.
• Proposed a RL method leveraging layer-wise freezing and a latent state replay buffer to enhance both sample and wall-time efficiency and wall-time in visual and continual learning tasks. Preliminary results show a 2x reduction in training time on ManiSkill3.

Research and Development Intern

Qualcomm Institute (Calit2)

Python, PyTorch, Unreal Engine, Kubernetes, Docker, C++

• Co-developing interactive 3D avatars of historical figures, driven by large language models, text-to-speech / animation pipeline, and Unreal Engine 5. Co-developing Climate Games, an educational video game, to raise awareness on climate change and archaeology. Both projects were presented at ASOR 2024 Annual Meeting.
• Developed a real-time audio-to-face pipeline that receives audio input from text-to-speech and uses NVIDIA Audio2Face through Rest API to animate facial movements on a 3D avatar. Further researched and developed our own multimodal co-speech gesture generation model for holistic body animation.
• Developed an Unreal plugin for real-time speech gesture generation and player communication. The plugin supports seamless communication in multiplayer gameplay and between Unreal Engine and external AI models.

Software Engineer Intern

Nearal

Kotlin, Android Studio

• Developed various features for Nearal’s Android application, including a dynamic onboarding screen, a floating login interface, and an improved sign-up process, using Kotlin and Android Studio.
• Refined multiple app fragments, ensuring optimal functionality for both logged-in and logged-out users, and resolved critical issues such as photo display inconsistencies and profile identification.

Instructional Assistant

UCSD Department of Computer Science and Engineering (CSE)

Instructional Assistant

UCSD Department of Computer Science and Engineering (CSE)

• CSE 152A: Taught Computer Vision and Deep Learning concepts for a class of 150+ students and assisted 20+ students weekly with programming assignments in office hours.

Regents Scholarship

UC San Diego

The most prestigious merit scholarship awarded to undergraduate students at the University of California.

USA Computing Olympiad - Platinum Division

USACO

Highest division in the most prestigious national pre-college algorithmic programming competition.