Kyle Zheng

Senior Research Scientist Ii @Cruise

US

I focus on research areas including machine learning, computer vision, multi-task learning, continuous learning, generalist, and generative AI. The goal of my team is to build an end-to-end off board vision-language-action driving model. My work includes both fundamental research to advance machine learning and targeted research to improve the performance of autonomous vehicles.I also work with machine learning infrastructure engineers on optimizing data and training pipelines. Our \'10x Training Project\' has significantly boosted model training speeds, achieving more than 10 times faster performance compared to previous implementations. The \'10x Data Pipeline\' initiative has also been effective in reducing costs by 39% and time by 27%, while maintaining the efficiency of model training.I work on leveraging the cruise.data, per-row-sampling, and other advanced MTL techniques to solve the long-tailed AV cases.I work on training the end-to-end trainable Vision-Language-Action (VLA) reasoning models on large-scale dataset so that the AV/ADAS systems can handle both common scenarios as well as long-tailed events with minimal human intervention.

Graduate University of Chinese Academy of Sciences

Master of Engineering (MEng)

University of Oxford

DPhil

Beijing Institute of Technology

Bachelor of Engineering (BEng)