Gordon Deng
About Gordon Deng
Gordon Deng is a Senior AI/ML Researcher and Technical Lead at Cerebras Systems, recognized for his innovative contributions to deep learning and meta-learning algorithms. He holds a Ph.D. in Information Engineering and has extensive experience in various roles, including positions at Huawei and PCI Geomatics.
Work at Cerebras Systems
Gordon Deng has been employed at Cerebras Systems as a Senior AI/ML Researcher and Technical Lead since 2021. In this role, he has contributed to the development of fast training algorithms and methodologies aimed at enhancing AI/ML model capabilities. His work focuses on optimizing deep learning models to achieve improved performance and generalization compared to existing methods.
Education and Expertise
Gordon Deng holds a Doctor of Philosophy (Ph.D.) in Information Engineering, specializing in image processing and pattern recognition, from Nanyang Technological University Singapore, completed between 1998 and 2001. He also earned a Master of Engineering (M.Eng.) in Remote Sensing Image Processing from the Research Institution of Petroleum Exploration and Development from 1992 to 1995. Additionally, he obtained a Bachelor of Science (B.S.) in Seismology and Geology from Peking University, studying from 1988 to 1992.
Background
Gordon Deng's professional background includes significant roles in various organizations. He worked as a Postdoctoral Fellow at the University of Waterloo from 2002 to 2004. Following this, he served as a Senior Scientist/Developer at PCI Geomatics from 2004 to 2018. He also held the position of Principal AI System Architect/Team Lead at Huawei Technologies from 2018 to 2021, and worked as an Information Technology Analyst at DBS Bank from 2000 to 2002.
Achievements
Gordon Deng has made notable contributions to the field of AI and machine learning. He innovated a meta-learning based training algorithm that won first prize in the ICLR 2021 competition. During his tenure at Huawei, he led the development of a GPU and Ascend-based prototype, resulting in one patent filed and one patent pending. He has also developed algorithms for simulating 3D convolution using 2D convolutions in video processing, enhancing both classification accuracy and computation speed.