Zhaoqing Wang (王兆卿)

I am a Research Scientist at Luma AI, with research interests in multimodal understanding and generation. My current work focuses on large-scale pretraining of world models, including dataset construction, model architecture, and training infrastructure optimization.

Previously, I was a Founding Research Scientist at PixVerse (AISphere), where I built the video generation training system from scratch and led base model training, dataset construction, and reinforcement learning fine-tuning, driving the development of PixVerse from V2 to V6. On the Artificial Analysis leaderboard, PixVerse V5 ranked #1 in Image-to-Video and #3 in Text-to-Video globally, while PixVerse V6 ranked #2 in Image-to-Video globally.

I received my MPhil and Ph.D. degrees from the Sydney AI Centre (SAIC) at the University of Sydney, supervised by Prof. Tongliang Liu and Prof. Mingming Gong.

My earlier research mainly focused on self-supervised learning and visual perception. I also completed research internships at Microsoft Research Asia with Wenlei Shi, OPPO Research Institute with Yandong Guo, Kuaishou Y-Tech with Qiang Li, and IDL of Baidu Research with Guodong Guo.