Celia Zhang

I’m a CS undergraduate student at Purdue.

Previously, I worked with Prof. Ahmed H. Qureshi at Purdue, and was a visiting researcher at BAIR at UC Berkeley, supervised by Prof. Jitendra Malik.

I always enjoy discussing and exchanging ideas across different areas of robotics, and I welcome conversations with anyone interested in the field, whether from academia, industry, startups, or investment.

Email  /  Google Scholar  /  LinkedIn  /  Github

profile photo

Research

My research focuses on foundation models for robotics. I am particularly interested in how large-scale training can endow robots with strong generalization and the ability to handle complex tasks. More concretely, my interests include leveraging pretrained models to help robots better understand the physical world, improving the data efficiency of imitation learning under limited demonstrations, and enabling closed-loop control through environment interaction.

Large Video Planner Enables Generalizable Robot Control

Boyuan Chen*, Tianyuan Zhang*, Haoran Geng*, Caiyi Zhang*, Peihao Li*, Kiwhan Song, William T. Freeman, Jitendra Malik, Pieter Abbeel, Russ Tedrake, Vincent Sitzmann, Yilun Du
In Submission
Website / ArXiv / HuggingFace

First open video model at a foundation-model scale for generative robotics planning.

Experience

LiberAI

Research Intern, Jan. 2026 ~ Apr. 2026
Worked on building robotics foundation model for bi-manual dexterous manipulation: Demo
Mentor: Fanqi Lin and Songming Liu
ZhenFund | 真格基金

Investment Intern, Dec. 2025 ~ Apr. 2026
Covered early-stage investments in AI and Robotics. Joined portfolio company LiberAI as a research intern concurrently.
Advisor: Emma Yin and Jiaqi Zhang
Berkeley Artificial Intelligence Research (BAIR) Lab

Research Assistant, Apr. 2025 ~ Nov. 2025
Advisor: Prof. Jitendra Malik
CoRAL Lab, Purdue CS

Research Assistant, Dec. 2024 ~ Mar. 2025
Advisor: Prof. Ahmed H. Qureshi

Selected Projects

Leap Hand VLA in Simulation

Rebuilt the DexGraspNet data collection pipeline in isaaclab to generate large-scale training data for dexterous hand manipulation. The pipeline supports VLA training across multiple model families including SmolVLA, OpenVLA, and the Pi series.
RoboVerse

A unified simulation and benchmarking framework for robot learning, enabling scalable data generation, cross-simulator interoperability, and consistent evaluation to advance imitation learning, reinforcement learning, and sim-to-real transfer.

Awards

  • Jandos Scholarship, supports students from underrepresented backgrounds in science, aligned with my belief in creating more opportunities for women and minorities in STEM, 2025
  • Dean's List, all semesters at Purdue
  • National First Prize, Chinese Mathematics Competition for College Students (CMC), 2022

Misc

Outside of research, I enjoy piano🎹, singing🎤 and reading📖, especially poetry.
I have been writing Chinese lyrics since middle school, evolving from early independent creations to contributing to soundtracks (OSTs) for radio dramas and TV series, with over 100M✨ total plays across platforms.
I have also been thinking about building a portfolio website for my lyrics, and would be happy to hear any thoughts or ideas :)

© 2026 Caiyi Zhang. Template adapted from Jon Barron.