Bojian Hou is an AI Research Scientist at Meta working on large-scale recommendation systems and trustworthy machine learning including calibration fairness and interpretability of large language models.
How media typically covers Bojian Hou
Bojian Hou as author
Kunlun, a new scalable architecture for recommendation systems, doubles scaling efficiency and increases GPU utilization from 17% to 37% by introducing optimizations like Generalized Dot-Product Attention and Computation Skip, now deployed in Meta Ads.
“Author of "Kunlun: Establishing Scaling Laws for Massive-Scale Recommendation Systems throu" on arXiv”