CV

Chenwei Wang

chenweiwang@connect.hku.hk
Pokfulam, Hong Kong, , HK, CN

Summary

Ph.D. Student

Education

  • Ph.D. student
    Present
    The University of Hong Kong
  • M.Phil. in Computer and Information Engineering
    Nov. 2025
    The Chinese University of Hong Kong Shenzhen
  • B.Eng. in Electronic Information Engineering
    June 2023
    Jilin University

Work Experience

  • Research Intern
    -
    Institute of AI China Telecom (TeleAI)
    • Developed a dual-autoencoder framework for efficient cross-robot fine-tuning using asymmetric Info-VAEs and a unified cross-embodiment action space.
    • Validated RDT policies in RoboTwin/ManiSkill3 and conducted preliminary deployment on a Unitree H1-2 humanoid with a RoboTera dexterous hand.
    • Designed a recursive action-averaging strategy for smooth and robust real-world control.
    • Developed a sparse-data VLA fine-tuning pipeline combining trajectory/video augmentation with real-world reinforcement learning.
  • Graduate Researcher
    2023 - 2024
    The Chinese University of Hong Kong Shenzhen
    • Built a fully autonomous ultrasound-guided biopsy system using a UR5e arm, ROS, U-Net++, 3D reconstruction, slice-to-volume registration, force/pose control, and reinforcement learning.
    • Proposed a robot-assisted quantitative compression elastography framework for accurate elasticity measurement.

Projects

  • Development of Data Collection System and Real-World Deployment Experiments on DP3 (Apr. 2025 – May 2025)
    Apr. 2025 – May 2025
    Developed data-collection tooling for the RoboTera dexterous hand and Unitree H1-2 humanoid, supported real-world DP3 deployment experiments, and contributed to paper writing.
  • Action Chunking Transformer Reimplementation and Demonstration at DTEC (Nov. 2024 – Dec. 2024)
    Nov. 2024 – Dec. 2024
    Contributed to teleoperation development and ACT/OpenTeleVision reproduction and adaptation; demonstrated the system at the TeleAI Developer Conference.
  • Spectrum-Sensing Chip for Explosive Detection (Dec. 2022 – May 2023)
    Dec. 2022 – May 2023
    Graduation project awarded Excellent Graduation Design by the School of Electronic Science and Engineering, Jilin University.
  • Vehicle Intelligent-Arm Grasping System Based on Motion Matching (May 2021 – May 2022)
    May 2021 – May 2022
    Team leader; awarded Outstanding Provincial Project.

Awards & Honors

  • Nominee for the Outstanding Graduate Student Award, The Chinese University of Hong Kong, Shenzhen, 2025
  • Top-Performing Intern, Institute of AI, China Telecom (TeleAI), 2025
  • Best Conference Paper Finalist nominee, IEEE RCAR, 2024
  • Excellent Graduation Design, School of Electronic Science and Engineering, Jilin University
  • Digital Technology Ecosystem Conference (DTEC), 2024 — Represented TeleAI and demonstrated embodied-AI and robotic technologies to academic, industry, and developer audiences.
  • World Artificial Intelligence Conference (WAIC), 2025 — Supported the preparation and coordination of TeleAI's exhibition and its on-site technology demonstrations.
  • Embodied Intelligence Hub (具身智能之心), 2025 — Presented TeleAI's robotic research in a livestream, demonstrating real-world embodied-intelligence systems and experimental results.
  • Jiqizhixin (机器之心), 2025 — TeleAI's embodied-AI and robotics work was featured in technical media coverage.

Publications

  • Robotic Assistance for Standardized Compression Ultrasound Elastography
    2024
    IEEE International Conference on Real-time Computing and Robotics (RCAR)
    A robot-assisted quantitative compression elastography framework for accurate elasticity measurement. Best Conference Paper Finalist nominee.
  • Enhancing Dexterous Diffusion Policy with Predictive 3D Representations
    2025
    Under review at IEEE Transactions on Cybernetics
    Under review at IEEE Transactions on Cybernetics. Chenwei Wang led the real-world robotic experiments.
  • Towards Fully Autonomous Robotic Ultrasound-Guided Biopsy for Superficial Organs
    2025
    IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)
    A fully autonomous ultrasound-guided biopsy system integrating image analysis, 3D reconstruction, registration, robot control, and reinforcement learning.
  • Align-Then-stEer: Adapting the Vision-Language Action Models through Unified Latent Guidance
    2026
    The Fourteenth International Conference on Learning Representations (ICLR)
    A data-efficient, plug-and-play framework for adapting vision-language-action models across tasks and robot embodiments through unified latent guidance.