Leo Lin

About me

Last updated: August 2026

I am a Computer Engineering student at the University of Illinois (UIUC). My interests include robot learning, computer vision, dexterous manipulation, and software engineering. My goal is to build robots that can work reliably in the messy real world. I'm currently doing robotics research under Lana Lazebnik and Shivansh Patel. I just wrapped up a robotics software engineer co-op at Amazon Robotics (Innovation Lab) under Prof. Taskin Padir.

Last summer, I interned at K-Scale Labs, training humanoid robots for manipulation with VLMs and action diffusion models, and writing infrastructure for data collection. I've also done research on robot arms, teleoperation, and computer vision through NSF REU and labs at UIUC.

At UIUC, I was a co-chair for SIGRobotics, a special interest group under ACM@UIUC, where I led the embodied AI hackathon team and humanoid project team. I'm a 3x robot hackathon winner, including 1st place locally at the LeRobot global hackathon and a dual-arm ninja shirt folding robot at Robot Hackathon NYC.

I'm graduating in December 2026 and looking for full-time roles in robotics. Feel free to contact me via email (leolinw07 at gmail dot com) or through any of the following links: LinkedIn, GitHub, Resume.

Publications

[NEW]CRAFT: A Tendon-Driven Hand with Hybrid Hard-Soft Compliance

Leo Lin*, Shivansh Patel*, Jay Moon*, Svetlana Lazebnik†, Unnat Jain†

RSS 2026
Best Paper Award — RSS 2026 Workshop on Dexterous Manipulation

paper | project website | BOM | code | bibtex 

@misc{lin2026crafttendondrivenhandhybrid,
  title={CRAFT: A Tendon-Driven Hand with Hybrid Hard-Soft Compliance},
  author={Leo Lin and Shivansh Patel and Jay Moon and Svetlana Lazebnik and Unnat Jain},
  year={2026},
  eprint={2603.12120},
  archivePrefix={arXiv},
  primaryClass={cs.RO},
  url={https://arxiv.org/abs/2603.12120},
}

Past work

Jan 2026 – Jun 2026 · Westborough, MA

Amazon Robotics — Software Development Engineer Co-op, Innovation Lab

Mapped SAM2-segmented textures and FoundationPose poses onto 3D primitives in Isaac Sim using Python and ROS2. Trained an Interactive World Simulator against NVIDIA's Cosmos world model to teleoperate a dreamed robot under high latency, and integrated Sharpa hands into Isaac Sim with Apple Vision Pro, Manus gloves, ROS2, and IsaacTeleop.

Jul 2025 – Aug 2025 · Palo Alto, CA

K-Scale Labs — Software Engineering / ML Intern

Implemented serial and UDP protocols to collect data for training humanoid manipulation VLAs, built real-time MuJoCo inference using a teleoperation setup, and fine-tuned GR00T-N1.5 to sort cubes based on language instructions on colors.

Jun 2025 · SF Bay Area · 1st place local

LeRobot Global Hackathon — Sock-sorting robot

Built a sock-sorting robot that won 1st place locally: Gemini for high-level planning, a teleop data-collection pipeline with bounding-box annotations, and VLA + ACT models trained with LeRobot's stack on cloud NVIDIA GPUs.

SF Bay Area LeRobot Hackathon winners with trophy

Research & projects