Profile photo

About

Starting January 2017, I will be an Assistant Professor at Carnegie Mellon University's Heinz College of Public Policy and Information Systems, focusing on machine learning for solving big societal problems! I will be looking for PhD students!

I'm currently a postdoc at MIT primarily working on online education with Polina Golland. I teach, develop, study, and apply machine learning and data analysis tools.

Prior to postdoc'ing and after five outrageous years in grad school at MIT, I staged a successful Ph.D. thesis offense on Star Wars Day 2015. My thesis was on machine learning for analyzing social data and medical images, and it won the George M. Sprowls award for best thesis in Computer Science at MIT! My advisors were Polina Golland and Devavrat Shah.

Before finding myself stranded in Massachusetts, I spent my childhood frolicking in the warmer climates of sunny California and completed my undergraduate studies at UC Berkeley in May 2010.

I enjoy teaching! Check out my teaching page for a list of classes I've taught at MIT, UC Berkeley, and in Jerusalem at a summer program MEET that brings together Israeli and Palestinian high school students. There are also videos of me teaching!

For a more formal bio, click here.

To hide my formal bio, click here.

Formal bio: George Chen is a postdoc at MIT in Electrical Engineering and Computer Science. In January 2017, he will join the faculty of Carnegie Mellon University's Heinz College of Public Policy and Information Systems. He teaches, develops, studies, and applies machine learning and data analysis tools. His work has spanned a diverse range of applications such as forecasting trends on Twitter, recommending products to people in systems like Netflix, finding human organs in medical images, and detecting buildings and villages in massive satellite images to help plan infrastructure development projects. George obtained his S.M. (2012), E.E. (2014), and Ph.D. (2015) degrees from the Electrical Engineering and Computer Science department at MIT, where he received the George M. Sprowls award for best Ph.D. thesis in Computer Science (2015). He previously completed his B.S. (2010) at UC Berkeley, dual majoring in Electrical Engineering and Computer Sciences, and Engineering Mathematics and Statistics. George enjoys teaching and has taught at the high school, undergraduate, and graduate levels, receiving teaching awards at UC Berkeley and MIT including the top graduate student teaching award at MIT, the Goodwin Medal (2015). His work with GridForm on analyzing satellite images to help bring electricity to rural India won the $10,000 grand prize at MIT's IDEAS Global Challenge 2014 for innovation and entrepreneurship in public service.

Current Projects

Here's what's on my plate:

New edX course on computational probability and inference
With faculty members Polina Golland, Greg Wornell, and Lizhong Zheng, as well as undergraduate Ali Soylemezoglu and recent graduate William Li, I'm developing an upcoming free online advanced-high-school/intro-college course on building computer programs that do probabilistic reasoning. The course teaches basic probability, probabilistic graphical models, and machine learning — all heavily intertwined with how to code things up! We're aiming to make this class accessible to anyone who knows introductory Python programming and calculus. Stay tuned!
Bringing electricity to those without it in developing countries
As part of GridForm, I analyze satellite images of enormous tracts of land to help bring electricity to rural regions in developing countries. Our work won the $10,000 grand prize at the MIT IDEAS Global Challenge (May 2014) and has been featured on Fast Company, SciDev, the Bangalore Mirror, and MIT News. I'm now also working with Stephen Lee, Bolei Zhou, and Claudio Vergara on automatically finding where buildings are in satellite images.
Nonparametric inference for analyzing social data and medical images
To forecast whether a news topic will go viral on Twitter, we can compare it to past news topics with similar Tweet activity. More generally, we can make a prediction based on an observation by looking at similar past observations. My Ph.D. thesis develops theory to understand when, why, and how well such so-called "nearest-neighbor" methods work in applications such as forecasting trends on Twitter, recommending products to people in systems like Netflix, and finding human organs in medical images. I'm doing some follow-up work during my postdoc.

My papers and past projects can be found here.

Miscellaneous

As a grad student, I moonlighted in various gigs:

Some of my other stints include racing for MIT's cycling team during the 2015 road season, supporting the Thirsty Ear student pub as a member of its executive committee (2013-2015), and handling a year of event publicity and photography for the EECS Graduate Student Association (2011).

I claim cycling, climbing, cooking, crafting libations, hitting the cinemas, skiing, and sewing as hobbies.


Last updated May 6, 2016. Photo credit: Danica Chang.