Fellowship Focus: A Conversation with Kartik Chandra
2021 Hertz Fellow Kartik Chandra applies tools built for machine learning systems in innovative ways to solve problems in visual computing.
A senior and honors candidate at Stanford University, Chandra is studying computer science, physics, and English literature. His undergraduate research has led to open-source software projects used by thousands of engineers. He has also built software tools for computer science education that are used by K-12 schools across the country.
On his current research:
My current interests are in taking tools we’ve developed from the past decade or so of deep learning and machine learning and applying those to study problems that aren’t necessarily machine learning related at all. We’ve got incredible software and algorithms that today are focused generally on deep learning, but in fact can be applied to all sorts of problems. I am particularly interested in using them to study human vision and visual perception.
On the challenges of machine learning:
I think one of the biggest challenges we as a research community are facing is to make the science we do understandable and trustworthy to the general public. We already have so many problems with machine learning systems being uninterpretable and having occasional strange or pathological behavior that can have real world consequences—really serious consequences—and it’s only going to get more and more urgent for us to look into solutions for this in the next 5 to 10 years.
On receiving a Hertz Fellowship:
I was very, very surprised. My immediate reaction was, what? Really? No way. But after that initial period of surprise, there was a feeling of being very humbled, receiving this award that so many colleagues, who have done such incredible work, have also received. Being part of this community of scientists gives me the opportunity to reach out to people who are working in different scientific fields. If I ever want to work on something interdisciplinary in a discipline where I have no connections or no prior experience, having this network of people is really valuable.
