UW-Madison Computer Sciences Department welcomes Sharon Li, assistant professor in deep learning and artificial intelligence

Sharon Li headshot

Sharon Li headshotSharon Li, one of Forbes 30-under-30- Science honorees, joins the CS Department from a post-doc at Stanford University. Li has been interested in math and physics since high school, and when she discovered machine learning, she found that it “connected beautifully with mathematics and various scientific disciplines.” Li works in deep learning and artificial intelligence and is particularly interested in exploring, understanding, and mitigating “the many challenges where failure modes can occur in deploying machine learning models in the open world.” Research in AI, such as the projects Li works on, fits the Wisconsin Idea perfectly, leading to “positive effects for people all over the world, like saving thousands of lives from auto accidents and helping every doctor diagnose and treat diseases earlier.” Li is particularly excited to be part of the new School of Computer, Data & Information Sciences, believing that being part of that “ecosystem” will lead to more collaboration and research opportunities. 

Hometown: Born and grew up in Lanzhou, a capital city in northwestern China. Historically, it has been a major link on the Silk Road. Lanzhou is well known for the hand-pulled beef noodles. The entire city has more than 1,000 beef noodle restaurants. To my biggest surprise, there is a restaurant right in Madison downtown offering the same kind of noodles!

Educational/professional background: I completed my Ph.D. from Cornell University in 2017, where I was fortunate to be advised by John E. Hopcroft. My thesis committee members are Kilian Q. Weinberger and Thorsten Joachims. Before joining UW-Madison, I spent a year as a postdoctoral researcher in the Computer Science Department at Stanford University, working with Chris Ré. Previously, I had also spent some time in the industry, where I worked as a Research Scientist at Facebook AI during 2017-2019. 

How did you get into your field of research? I have always liked math and physics since high school. My undergraduate education in engineering opened up the door for me to use mathematics to solve practical problems. When I started my Ph.D. at Cornell University, I took a class on the Mathematical Foundations of Data Science, which was taught by Prof. John Hopcroft. The course was my first exposure to machine learning and AI. I was fascinated by the idea of building a mathematical model to make powerful predictions in the real world. I found machine learning connected beautifully with mathematics and various scientific disciplines. It also aligned with my passion for seeking to explore, understand, and solve real-world problems. Later I decided to pursue my Ph.D. under John’s advisory and concentrate on the area of deep learning (which is a subfield of machine learning) for my thesis research.

Could you please describe your area of focus? My broad research interests are in deep learning and its applications. My time in both academia and industry has shaped my view and approach in research. Particularly, my research focuses on uncertainty and robustness in deep neural networks. My work explores, understands, and mitigates the many challenges where failure modes can occur in deploying machine learning models in the open world. The goal of my research is to enable transformative algorithms and practices towards open-world machine learning, which can function safely and adaptively in the presence of evolving and unpredictable data streams. 

What main issue do you address or problem do you seek to solve in your work? One of the issues I am working on is uncertainty estimation and out-of-distribution detection, which is a critical component of building open-world AI systems. The problem is important because the classic closed-world assumption, i.e., both the training and test data are drawn from the same distribution, rarely holds in modern deep learning applications. In contrast, the real world is open and full of unknowns. Determining whether inputs are out-of-distribution is an essential building block for safely deploying ML in the open world, especially for safety-critical applications such as biometric authentication, autonomous driving, and medical diagnosis. 

What’s one thing you hope students who take a class with you will come away with? The enthusiasm for using artificial intelligence to make a positive change in the world and the essential knowledge to achieve that goal. 

What attracted you to UW-Madison? The Computer Sciences Department at UW-Madison is very strong nationwide and has a spectacular set of colleagues in different areas. I am thrilled to communicate and collaborate with them. I had a great interview experience and was quite impressed by the collegial and intellectually stimulating atmosphere in the department. In addition, the newly established School of Computer, Data & Information Sciences (CDIS) is a great ecosystem to be in. It provides a unique opportunity to create collaborative educational and research programs across various disciplines. I am excited to be part of the initiative and advance my career here. Lastly, as a person always attracted to the lake and ocean (as evidenced in places I’ve lived), how could I not fall for the beautiful Lake Mendota and Lake Monona? 

What was your first visit to campus like? My first visit was for the job interview back in March 2019, and it had twists and turns. The night before my interview, I had to evacuate from the South Union (where I stayed) due to a midnight fire alarm. I frankly didn’t think I would make it through the job interview. But thankfully, people here are warm and intellectually stimulating enough to compensate for all of that sleepless night. After the interview, my faculty host kindly showed me around the campus along the shore of Lake Mendota (and it was fun to see ice fishing for the first time). The university is grand looking and reminded me of Cornell in certain ways. It’s almost hard to believe, but I already had a sense of home and was really looking forward to coming back to Madison in the summertime. And now here I am! 

What are you looking forward to doing or experiencing in Madison? There are many things on my bucket list: spend an afternoon reading by the shore, explore the Chazen Museum, bike loop around the lakes, revisit the zoo, and shop at the local farmer’s market (when it reopens).  

Do you feel your work relates in any way to the Wisconsin Idea? Artificial intelligence has been transforming our daily lives in unprecedented ways, especially with the new scientific advancements. Enabling trustworthy open-world machine learning helps keep a beneficial impact on society. The research can lead to positive effects for people all over the world, like saving thousands of lives from auto accidents and helping every doctor diagnose and treat diseases earlier. I think this effort is very much in line with the Wisconsin Idea.

Hobbies/other interests: Outside work, I enjoy photography, traveling and watching stand-up comedy.