Heng Zhang
AI Technology CTO at Guoxing Aerospace, former Senior Algorithm Expert at Xiaomi
He is currently the AI Technology CTO of Guoxing Aerospace. He received his Ph.D. in Computational Biochemistry from Imperial College London in 2018. He has published more than 10 papers as first author or co-author in international journals and top conferences such as CVPR and ICCV, and holds over 15 patents. His research interests span large-scale digital twins, generative AI, multimodal large language models, low-level vision enhancement, and next-generation image signal processing systems. From 2023 to 2025, he led his team to win a total of five championships in CVPR competitions across two consecutive years. He excels at translating academic research into mature algorithmic products, driving the deployment of vision algorithms on edge devices. He has been awarded a Marie Curie Fellowship by the European Commission and a MIT–Imperial College Global Fellowship. He was recognized as a Shenzhen Overseas High-Level Talent, and since 2024, he has served as Chair Professor of Artificial Intelligence at Beihang University.
Topic
Research on the Industrialization of AI Visual Algorithms — Full-Chain Exploration from Academic Frontiers to Industrial Practice
This report focuses on the critical path from R\&D to commercialization of AI vision algorithms, drawing on the speaker’s decade-long academic background in computer vision and hands-on experience at leading tech companies. It systematically outlines core methodologies for translating state-of-the-art algorithms from top-tier conferences like CVPR and ICCV into real-world applications such as mobile imaging and smart hardware. Topics include solutions to engineering challenges such as data pipeline optimization, lightweight model deployment, and cross-modal fusion technologies. The report will take case studies from the speaker’s leadership in AI imaging algorithms at OnePlus and multimedia AI product development at Xiaomi to dissect the full-cycle development process—from lab prototypes to products used by hundreds of millions. It will explore strategies for balancing algorithm performance with user experience. Additionally, it will highlight over 20 patented innovations and five championship-winning CVPR competition techniques to illustrate the collaborative innovation model between academia and industry. The report also discusses deployment trends and ethical considerations for AI vision in areas such as smart hardware and intelligent cities. This talk is intended for AI algorithm engineers, product managers, and researchers in related fields, offering insights into building a “Technology–Product–Scenario” integrated approach to real-world implementation.
