cv

Basics

Name Mengying Lin
Label MS Robotics @ Georgia Institute of Technology
Email mlin365@gatech.edu
Url https://Cassie-Lim.github.io/

Work

  • 2024.12 - Present
    Student Researcher
    Robot Learning and Reasoning Lab, Georgia Institute of Technology
    Developed a part-level graph correspondence framework that enables zero-shot manipulation transfer across unseen object categories. Designed multi-scale object representations and an optimal-transport–based matching algorithm to propagate grasps and trajectories directly from geometric structure. Integrated the method with MimicGen to generate diverse demonstrations and validated its effectiveness through simulation and real-world execution on a Franka Panda robot.
  • 2024.03 - 2024.10
    Research Intern
    Institute for AI Industry Research, Tsinghua University
    Developed a task-aware neural field system that enabled more accurate scene understanding and supported robust multi-modal localization and navigation. Designed a selective experience replay strategy to prioritize task-relevant, high-confidence features, effectively reducing semantic noise during continual learning. The resulting system achieved a +4.36 improvement in localization accuracy over the previous state of the art, significantly enhancing downstream navigation reliability.
  • 2023.08 - 2024.9
    Student Researcher
    Institute of Automation, Chinese Academy of Sciences
    Developed a plug-and-train module that dynamically balances LLM-derived general priors with dataset-specific priors to improve downstream policy learning. The approach achieved state-of-the-art performance in both seen and unseen environments in the AI2-THOR simulator and successfully transferred navigation policies to physical robots in a zero-shot setting, demonstrating reliable real-world task completion. (IROS 2025 oral.)

Education

Skills

Robotic Navigation
Visual Perception
Robot Operating System (ROS)
Simulation
Git
C++
Python

Languages

Chinese
Native speaker
English
Fluent