Education

  • 2023 - present
    Ph.D, Computer Science, UC Berkeley
    • Ph.D. student advised by Professor Sergey Levine. Please see my research page for details.
    • Selected courses:
      Deep Reinforcement Learning Computer Vision Machine Learning Systems Convex Optimization
  • 2019 - 2023
    Sc.B., Applied Mathematics & Computer Science, Brown University
    • GPA: 4.0 / magna cum laude / CS Honors / Sigma Xi
    • CS courses:
      Advanced Deep Learning (grad level) Machine Learning Computer Vision Collaborative Robotics (grad level) Computer Systems Software Engineering Multiprocessor Syncrhonization
    • Applied Math courses:
      Recent Applciations of Computational Probability and Statistics Pattern Theory Honors Statistical Inference Computational Linear Algebra Applied ODE/PDE Honors Calculus and Linear Algebra

Honors and Awards

    • Best Paper Award at RSS 2025 Robot Evaluation Workshop
    • Berkeley College of Engineering Fellowship
    • Brown University magna cum laude, CS Honors, CS Senior Prize, Sigma Xi
    • Placed 227th (top 5%) in Putnam 2019, top 3 at Brown
    • 2nd Place in Hartshorn-Hypatia Math Contest
    • Brown UTRA research scholarship
    • only recipient of Yongren Full Fellowship at PROMYS (2018)
    • Provincial Top 1% in Chinese Physics Olympiad
    • Regional Top 10 & International Top 100 in Physics Bowl
    • Finalist in High School Mathematical Contest in Modeling (HiMCM)

Experience

  • May 2025 - Present
    Research Intern, Physical Intelligence
    Robot Learning VLA Reinforcement Learning
    • Led the development of advantage-conditioned Vision-Language-Action (VLA) models, which culminated in the new flagship model π*0.6, the first large-scale VLA trained with offline RL.
    • Researched pre-training and post-training of advantage-conditioned VLAs, and increased task success rate to near 100% and doubled the task throughput. We enabled the robot to autonomously make coffee for 13h without failures.
    • Trained the first multi-task value functions over large internal datasets to identify good and bad parts of the dataset, and worked with the team to use the value function for policy improvement and learning from failures.
  • Jan 2021 - May 2023
    Undergraduate Researcher, Brown University
    Research Reinforcement Learning Machine Learning/AI Python Pytorch
  • Jan 2022 - May 2022
    Head Teaching Assistant, Brown University
    Autograder Python Management Course Development
    • Managed a team of 20 teaching assistants and organized the course logistics for 200 students, and handled communication between the professor, teaching assistants, and students.
    • Built auto-grading pipeline for 12 coding assignments on Gradescope and enabled students to see code correctness shortly after handin.
    • Answered questions through weekly TA hours and the online discussion platform Edstem.
  • July 2021 - Aug 2021
    Natural Language Processing Engineer Intern, Zencastr Inc.
    Machine Learning NLP MongoDB Python C++
    • Engineered and deployed a web app with websockets and FastAPI that allows users to edit (faulty) audio-to-text automatic transcriptions and provides a faster editing experience by intelligently recommending potentially incorrect segments: the recommendations are made by finding similar occurrences of user-made edits throughout the audio file with Keyword Spotting using language and acoustic models from Kaldi and Vosk-api.
    • Sped up Keyword Spotting 2x using multithreaded offline-decoding in Python and Shell; sped up automatic speech recognition 5x using WeNet architecture (written in C++) and Speech Activity Detection models from Kaldi; model is pushed to production.
    • Implemented a thread-safe MongoDB store with asyncio and motor to store user-made edits in the backend.
  • Dec 2020 - Jan 2021
    Machine Learning Engineer Intern, Zencastr Inc.
    Machine Learning NLP data augmentation CNN Keras Python
    • Built a CNN in Keras that classifies audio files into speech, music, laughter, or noise with 93% accuracy; trained using audio data crawled from YouTube using youtube-dl and augmented by adding noise, changing pitch, and stretching time.
    • Aligned audio-to-text transcriptions from DeepSpeech and Webspeech API using dynamic time warping and grapheme confusion.
    • Built a private Python package of machine learning utility scripts hosted on GitHub with Continuous Integration.

Techinical Skills

  • Python PyTorch TensorFlow/Keras C Docker MongoDB Java JavaScript React MATLAB Assembly x86-64 Scala ReasonML C#