Summary
Chengcheng Ma is a research scientist with a PhD from the Institute of Automation, CAS, and about a decade of experience advancing vision and generative AI across leading Chinese tech labs and industry teams. He specializes in text-to-video and text-to-image generation, semi-supervised and long-tail learning, and adversarial robustness, with applied work spanning world models, embodied AI, and practical dataset curation (e.g., improving LAION captions). His publications include methods for mitigating prompt-tuning overfitting, complementary experts for long-tailed semi-supervised learning, and robust adversarial detection, reflecting a balance of theoretical insight and engineering impact. Chengcheng has transitioned smoothly between research internships at Tencent, Huawei, and others to research roles at Kunlun and Alibaba, demonstrating an ability to ship reproducible results and deployable techniques. He maintains an active academic profile on Google Scholar and an open-source presence on GitHub, signaling commitment to both open research and production-ready systems.
10 years of coding experience
2 years of employment as a software developer
Bachelor of Engineering - BE Automation, Bachelor of Engineering - BE Automation at Northwestern Polytechnical University
Doctor of Philosophy - PhD Pattern Recognition and Intelligence System, Doctor of Philosophy - PhD Pattern Recognition and Intelligence System at Institute of Automation, Chinese Academy of Sciences