Moy Yuan

I am Moy (Zhangdie) Yuan, an Applied Scientist at Amazon AGI, where I work on the post-training and alignment of multimodal foundation models. My work contributes to the Nova family of models, including the audio-centric Nova Sonic, and powers next-generation experiences for Alexa+ which serve millions of users worldwide.

Beyond building capable voice interfaces, my research focuses on the fundamental reasoning capabilities of these systems. I am particularly interested in developing World Models that can accurately represent and anticipate changes in the real world. This involves moving beyond static benchmarks to predict unfolding events while ensuring outputs remain factual and grounded in reality.

I recently finished my PhD thesis on Language Models as Forecasters and am awaiting my viva at Cambridge, advised by Andreas Vlachos. My PhD was generously funded by the ERC and Trinity Hall. Before Cambridge, I completed my BSc in Computer Science from Edinburgh.