Dr. Adriana Kovashka’s NSF CAREER project aims to lower the cost of AI learning

Dr. Adriana Kovashka received a National Science Foundation (NSF) CAREER award in May 2021. The Faculty Early Career Development (CAREER) Program is an NSF-wide activity that offers the Foundation's most prestigious awards in support of early-career faculty who have the potential to serve as academic role models in research and education and to lead advances in the mission of their department or organization.

Dr. Kovashka, who is originally from Bulgaria, teaches deep learning at the undergraduate level, and several graduate courses, including computer vision, both of which involve forms of machine learning. Her latest project, for which she was awarded the CAREER grant, focuses on object detection in images. 

According to Dr. Kovashka, the ultimate goal of her research is to lower the cost of AI learning for future researchers. While the current methods require humans to teach computers how to recognize objects by annotating millions of images, Kovashka hopes to reduce the cost of this sort of machine learning by teaching systems to learn and recognize objects without as much human input. 

“These computer vision systems could be trained, maybe ten years down the line in a more natural way,” she said. “You [could] have a robot that knows nothing when it starts but — by living with you and seeing you interact with other people —  it can ultimately pick up on what these words mean and what objects are good for and what they can be used for.”

Dr. Rebecca Hwa, a professor in the CS department and one of Kovashka’s collaborators and mentor, explained that the NSF CAREER award is unique in that it is often viewed as a milestone for junior faculty and a way for recipients to receive accolation from the broader research community, no matter what their discipline. 

“The career grant is unlike a typical research grant from the NSF in that it is about setting the foundation of this particular researcher,” she said. “It’s supposed to show that you are going to be the expert in this particular research direction that you propose and it’s a direction that your fellow researchers find value in.”

In addition to her research, Kovashka hopes to use her lab as a way to bring the world of computer science to people who may not have access to as many resources as are available at Pitt CS. According to Kovashka, a majority of computer science outreach is focused on mentoring undergraduates from one’s own institution, and she seeks to change that.

“[Pitt students] are probably in a better place along their educational path than students in, say, community colleges, in terms of having access to computing research,” she said. “I would be interested in bringing people from community colleges, which are smaller and maybe have smaller computer science programs to do research in my lab.”

Mingda Zhang, a fifth year PhD student in Dr. Kovashka’s lab said that he’s experienced Kovashka as not just an excellent mentor professionally and academically, but as part of a support system for him while he completes his degree abroad. 

“During the COVID pandemic as an international student I am in the United States to pursue my degree, but all of my family is in China. My grandmother passed away last year, and it was a very difficult time,” he said. “[Dr. Kovashka is] super understanding and super supportive and helped me go through that difficult time.”

In addition to the courses she currently teaches, Dr. Kovashka started a course this semester for graduate students focusing on writing skills. Stemming from her own difficulties as a graduate student, Kovashka hopes to set aside time to help current students hone their skills before they’re in a time crunch with paper deadlines. 

“Everyone mentors their own graduate students and teaches them to write better but it usually happens before deadlines when you don’t have a lot of time, so having a dedicated course where we can specifically talk about how do you argue your case and how do you motivate your case, I think would be something that grad students would benefit from,” she said.

As an undergraduate at Pomona College, Dr. Kovashka majored in media studies as well as computer science which informed some of her earlier work in the department studying the rhetoric of visual advertisements. Her background in the arts also influenced her decision to come to Pitt when she received an email about Pittsburgh’s thriving film scene from Dr. Jan Wiebe, who had a feeling Kovashka was a film buff. 

“I was really impressed by that, that A) there's people who are so perceptive and B) that, you know, there was this culture and these film festivals, and all that, so that was cool and those are some of the reasons why I chose Pitt over other options.”

As part of Pitt CS, which Kovashka describes as having a vibrant environment and community, Kovashka has had the opportunity to work with faculty not just from her own department, but from others, such as psychology, as well. Dr. Rebecca Hwa said she admired Dr. Kovashka’s ability to organize and write a proposal that had a multidisciplinary approach and involved the work of many different faculty.

“Adriana was able to run the meetings efficiently and get us all to say what is helpful for the proposal but without cutting anybody off or anything like that,” she said. “I thought that was really well organized and it’s the sort of people skill that you have to be fairly experienced to be able to carry out.”

According to Dr. Hwa, who has been partnered with Dr. Kovashka and was her faculty mentor since she first joined the department in early 2015, the NSF CAREER grant is just the latest example of Kovashka’s excellent work as both a researcher and a mentor. 

“It’s very, very rewarding to see how well she grew into this super capable researcher,” she said. “I’m actually really impressed by everything she does and, you know, to think in a way [as Dr. Kovashka’s mentor] I am the lucky one.”

To hear more from Dr. Kovashka and about her research, register for her talk on October 19th, 2021 as part of the new SCI Dean's Department and Program Spotlight Series


Research Areas

News Type