Bojian Hou

My research has moved from learning theory to machine learning at production scale, with one thread running through it: making models trustworthy, so that they stay robust, fair, interpretable, and well calibrated in the settings where they are actually used.

I earned my B.Sc. and Ph.D. in Computer Science at Nanjing University, where my doctoral work with Prof. Zhi-Hua Zhou in the LAMDA Group studied learning in open environments: data arrive as streams, and features appear, disappear, and evolve over time. This line of work appeared at NeurIPS, IJCAI, and AAAI and in TKDE and TNNLS, and received excellent doctoral dissertation awards from Jiangsu Province, JSAI, and Nanjing University.

As a postdoctoral fellow in the Department of Biostatistics, Epidemiology and Informatics at the University of Pennsylvania, advised by Prof. Li Shen, I led a Trustworthy AI research line on fairness, interpretability, and calibration, from classical models to large language models and often with medical data where errors are costly. I mentored 20+ students there, and our work appeared at NeurIPS, ICML, ICLR, and KDD.

At Meta, I bring the same perspective to recommendation at production scale, caring about both the theoretical foundations and the real-world deployment of machine learning.