About Me

I’m ZhenTing Liu (刘镇霆) or Jerry, currently a sophomore at Carnegie Mellon University. I am interested in optimization theory, robotics (mostly UAVs), algebra, geometry and deep learning . My current research focus includes differentiable manifolds, multi-agent collaboration and optimization. My ultimate goal is to obtain a PhD in above area. In the meantime I also run a few non-profits (Voice link invalid temporarily as I am migrating the domain; Airsave).

I worked closely with the Sri Lankan government and researchers on local water treatment projects using microbial fuel-cells, which I included in the blog post

I did semi-professional bouldering in high school and ref for regionals when I have time. Here are some of my past competitions. I love traveling, I took a year off to backpack across south America and Asia while doing non-profit work. I have a blog post about this.

I enjoy playing the piano, photography, golfing and snowboarding. I have a Motor Glider license and am working toward getting a FAA Sport Pilot license. I played a lot of video games in high school, especially this pretty niched RTS game where I was ranked top 3 worldwide for a few years. I love Gomoku too although I’m not very good at it.

I also post logic (and some other) puzzles here, check them out!

Feel free to reach out to me at zhentinl[at]andrew[dot]cmu[dot]edu

Research Experience

AirLab                                                                  Sep 2024 - Now

Multi-robot collaboration and optimization

SURF                                                                  May 2024 - Now

SLAM for UAV Swarm and State Control using Multi-Framework System
Using decentralized and distributed framework for multi-UAV control and pose-graph optimization.

ORI                                                                  Dec 2020 - May 2021

Learning to Learn Semantic Factors in Heterogeneous Image Classification
Proposed Meta-ProtoNet to handle the challenge of heterogeneous task distributions in few-shot scenarios. Conducted experiments on 5-way 1-shot and 5-way 5-shot settings with 15 query samples per class in each. Result demonstrates that Meta-ProtoNet outperforms the original Prototypical Network in all conventional FSL scenarios.

University of Oxford                                                                  Dec 2020 - Feb 2021

Research Intern (under the supervision of Professor Alex Rogers)
Build ML models to minimize household energy expenditure in the UK through proposing different usage strategies. Implemented algorithms for Joulo, a home energy advice system using a low-cost temperature sensor, on cloud.

Independent Work                                                                  May 2020 - Sep 2020

GNNs for graph-based semi-supervised learning(SSL)
Explored GNNs for soft SSL in which vertex labels are probability distributions. Motivated by real-world applications, we proposed DHN (Directed Hypergraph Network), a novel method for directed hypergraphs. DHN can effectively propagate histograms by integrating directed hyper-edges and undirected hyper-graphic structure into unknown vertices of the vertex features.

Work Experience

• ML engineer intern @Tesla in 2020
• Frontend @Matterverse in 2021