Magazine article American Scientist

Meet Your Fellow Companion: Junfeng Pan

Magazine article American Scientist

Meet Your Fellow Companion: Junfeng Pan

Article excerpt

We are pleased to feature Junfeng Pan, currently a research scientist and tech lead for ads data and model engineering at Facebook in Menlo Park, California, in this issue's Meet Your Fellow Companions Interview Series. After completing extensive coursework for his master's and Ph.D. in computer science in China, Pan moved to the United States to work first for Google and now Facebook in the ever-expanding world of data mining and machine learning. A dedicated Sigma Xi member, Pan's current work in artificial intelligence constantly requires interdisciplinary science knowledge and teamwork. He is an outstanding example of a fresh face of Sigma Xi's future and we are proud to feature him here in our Meet Your Fellow Companions Series.

What is the focus of your current research?

Currently, my main focus is on maximizing the user and advertiser experience by finding the best matches of ads and users from user responses. It is a complex smart learning model and system that can learn on large scales and in nearly real time. It simulates the human learning process at a certain level. For example, when we learn to play a song on the piano, there is an easy and a hard pan. When we make a mistake, we focus on the challenging pan and practice to improve. Our learning system is similar to this learning process as it moves through the stages of prediction, actual user feedback, recognition of the error, then improvement of the prediction in real time. The difference is that, our model and system work at scale. It is a distributed system spread out over a huge set of machines in multiple global regions. It learns and predicts in the magnitude of billions or trillions on a daily basis.

Tell us about something we might see in our daily lives that directly correlates to your work.

This is the most straightforward question I can answer. If you are a Facebook user, you see my work. Facebook has over 1 billion monthly active users, forming the largest scale social network on the planet.

Give us an example of how multi-disciplinary research directly contributed to your work.

This is the most exciting part of my work, as it crosses at least three domains. First, I work on real-time learning and prediction models and systems, which are based on a solid foundation of mathematics, probability and artificial intelligence, all of which aim to simulate human learning. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed


An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.