04 — Experience

Experience

Internships and education, most recent first.

Internship experience

Jun 1 – Aug 31, 2026

VLA Algorithm Engineer (Intern)

Honor Device Co., Ltd.Shanghai, China

  • Owned a human-in-the-loop real-robot learning stack for VLA models on Honor’s self-developed Omega humanoid, end to end across six parts:
  • End-to-end VLA pipeline: set up the lab’s collect → post-train → deploy loop from scratch (two teleoperation rigs, operator training, GR00T N1.7 post-training, chunked real-time inference).
  • Offline data design: four demonstration categories (standard, pick and place perturbations, failure recovery) so the policy re-tracks moving targets and recovers from missed grasps.
  • Human-takeover training infra: smooth teleoperator takeover, an intervention-driven data pipeline and a distributed Actor–Learner–Robot loop (AGX/NX, x86 Actor, GPU Learner); a mirrored-layout task rose from near 0% to 50% autonomous success in about 140 episodes.
  • Real-robot RL algorithms: ported RECAP (π0.6) advantage-conditioned training to GR00T; designed RAPID, a DSRL-style noise-space adaptation that reuses human corrections before a new checkpoint lands; extended the stack to loco-manipulation.
  • ROVE reproduction: reproduced XPENG’s ROVE on fine-grained phone grasping — stage-aware returns plus an optimistic (expectile) value function that pins negative advantage on takeover-causing actions; autonomous success from 40% to 52%.
  • Real-robot debugging: traced six system-level bugs (colour-space mismatch, classifier hysteresis, yaw drift, stale camera frames) with quantitative probes.
  • Result: an in-house loop in which the robot learns new tasks from human takeovers — spanning data, systems, RL algorithms and deployment.
Mar 2026
— May 2026

Research Intern

Huawei Norbert Wiener Research CenterSingapore

  • Developed backend deep learning models for short-range radar signal processing.
  • Conducted research on downstream tasks including human pose estimation and indoor mapping/SLAM from radar data.
Dec 2025
— Jan 2026

Robotics Software Intern

Movel.AISingapore

  • On-site validation and factory deployment of autonomous navigation for forklift robots in real warehouse environments (ROS2).

Education

Aug 2025
— Present

Master of Science in Electrical Engineering

National University of SingaporeSingapore

  • Core member of the NUS CORE Lab (COntrol, Robotics & Artificial IntElligence), focusing on Embodied AI and autonomous robotic decision-making.
  • Research on perception and planning for aerial robots based on Vision-Language Navigation (VLN), and end-to-end Reinforcement Learning in complex 3D environments.
Sep 2021
— Jul 2025

Bachelor of Engineering in Electronic and Communication Engineering

Nanjing University of Aeronautics and AstronauticsNanjing, China

  • GPA: 88/100
  • Key Courses: Advanced Mathematics (96), Probability Theory (97), Random Signal Analysis (97), Microwave Technology and Antenna (97), Fundamentals of Communication (95), Complex Function (99).