Yunzhe "Jonas" Li
Exciting updates coming soon!
My professional experience centered on advancing the autonomy frontiers of perception and scene understanding at scale. At WeRide, I contributed to the core model of WePilot AiDrive system by owning detection suites and developing a profiler for 900 million training data points. I also pioneered a "Chain-of-Causation" annotation schema for Vision-Language Models fine-tuning to bridge the gap in understanding driving intent through behavior captioning. Prior to that at Momenta, I delivered the initial autopilot reversing feature for the GM Cadillac LYRIQ, successfully reducing false positives by 15% on 800+ real parking lot cases across 30+ garages.
Prior to the industry, I earned my Master's in EECS at UC Berkeley in 2025 -- Fiat Lux and Go Bears, following my B.Eng. in Computer Science from Shanghai University in 2024. My passion for the intersection of computer vision, AI, and robotics began in high school, which encouraged me to continue my journey in this field through undergraduate, master, and the industry.
In RoboMaster competition from 2020 to 2024, I co-led the computer vision subteam and later served as overall lead of a 40-member squad, where we built eight types of robots from scratch and won 3rd place in the Shanghai Regional Competition. From 2017 to 2020, I led a team in the FIRST Tech Challenge. We earned 2 FIRST World Championship qualifications, 1 Inspire Award, and 4 Connect Awards. I am graciously to be an alumnus of both RoboMaster and FIRST, and affiliated with DJI as an event technical executive.
