Summary
Chang Zeng is an AI researcher and PhD candidate based in Chiyoda, Japan, with 8 years of experience in speech signal processing, sequence-to-sequence models, and deep learning across academia and industry. He has driven high-impact speech projects—from large-scale speaker recognition and live-broadcast monitoring at Alibaba to full-duplex conversational systems at 理想汽车—and has published over a dozen papers at venues like Interspeech and ICASSP. His research portfolio includes singing voice and TTS systems (e.g., XiaoiceSing2, CrossSinger) and GAN-based vocoders, with work that enabled high-fidelity, zero-shot multi-singer synthesis. Now focused on deep generative AI for language and audio, he bridges rigorous academic research with production deployment and is building toward a long-term career in generative modeling. An understudied strength is his track record of turning research prototypes into deployed systems that meet real-world scale and regulatory needs.
8 years of coding experience
5 years of employment as a software developer
Doctor's Degree, Informatics, Doctor's Degree, Informatics at National Institute of Informatics
University of Tokyo
Bachelor's degree, Measuring and Controlling Technologies and Instruments, GPA 3.74, Bachelor's degree, Measuring and Controlling Technologies and Instruments, GPA 3.74 at 天津大学
Doctor of Philosophy - PhD, Informatics, Doctor of Philosophy - PhD, Informatics at The Graduate University for Advanced Studies, SOKENDAI
Japanese, Chinese, Chinese