IDEaS is one of Georgia Tech's 10 interdisciplinary research institutes within the Georgia Tech Research enterprise.

What is your field of expertise and why did you choose it? 
Growing up, I liked design as a hobby from designing posters to making websites, and I also liked engineering and computer science. As I was close to graduating with my bachelor's degree in Hong Kong, I discovered the field of human-computer interaction or HCI. It's now well known, but not so much in the days when I was a student. HCI intersects computing, design, and psychology — exactly what I wanted to learn more about for my master’s degree at Carnegie Mellon. My machine learning Ph.D. thesis research intersected data mining and HCI, helping people make sense of large network datasets. Now as a faculty member, my group’s core research vision echoes a similar focus, taking the best of both worlds to combine what people and machines do best. I now work at the intersection of machine learning and visualization. My group innovates scalable, interactive, and interpretable tools that amplify a human's ability to understand and interact with billion-scale data and machine learning models. Our current research thrusts include human-centered AI (interpretable, fair, safe AI); large graph visualization and mining; cybersecurity; and social good.

What makes Georgia Tech research institutes unique? 
Their breadth of the strategic areas that they cover and the researchers that they involve across the whole campus’s colleges and departments. The institutes are great reflections of the huge number and variety of talents at Georgia Tech.

What impact is your research having on the world?
My group has open-sourced and deployed many systems and tools over the years. Some recent examples include CNN Explainer and GAN Lab (with Google Brain): a suite of accessible tools for students and experts alike to learn about AI models and now used in courses around the world. They also include ActiVis, a first visualization for interpreting industry-scale AI models, and Summit and Bluff, systems that scalably summarize and visualize what features a deep learning model has learned, how those features interact to make predictions, and how they may be exploited by attacks. There’s also MalNet, the largest public cybersecurity database with more than 1.2 million graphs and images, and SHIELD, a fast defense that removes adversarial noise by stochastic data compression.

What do you like to do in your spare time when you are not working on your research or teaching?
Painting simple abstract art and playing the cello and piano.