Whether you are aware of it or not, a good portion of what you see when you are online is probably dictated by recommendation systems. These systems customize our internet experience by predicting what videos, music, articles, and products we will be most interested in, and can sometimes be even more accurate than our friends in knowing what we will and won’t like. Alumna Pigi Kouki (’18) spent her years as a Ph.D. student at Baskin Engineering studying these types of recommendation systems, and now she is implementing them as a data scientist at RelationalAI, a California-based company that seeks to push the boundaries of artificial intelligence.
Kouki came to the Baskin School of Engineering at UCSC from the University of Athens, where she had completed both a B.S. and an M.Sc. degree in computer science. She already knew that she wanted to focus her graduate research on data mining and recommender systems, so Lise Getoor’s LINQS Statistical Relational Learning Group (which focuses on statistical relational learning, probabilistic reasoning and scalable inference, and their applications in user modelling, explainable recommender systems, and social good), was a perfect fit.
“During my Ph.D., I was very fortunate to work in problems that are of high interest to both academia and industry”
“My research mainly focused on entity resolution, recommendation systems, and explanations,” Kouki explains. “I focused on implementing algorithms that can come up with recommendations that are relevant to the users, such as what movie to watch next.”
Kouki was also interested not just in making recommendations, but also in explaining to users how those recommendations were made. “It is important to also explain to the users the reasoning that the algorithm followed in order to come up with a particular recommendation,” she says. “As users become more familiar with recommendations in their everyday life, the need to know how the system works is becoming more evident.”
The LINQS lab was a tight-knit group, and Kouki says that she has fond memories of the lunches they would hold to celebrate members’ accomplishments, and the walks they would have in the redwoods on the UCSC campus. She expresses gratitude for the collaborations she was able to have with fellow grad students and postdocs, and especially for her advisor, Professor Lise Getoor, who she says was an excellent mentor to her as a woman in tech.
“Lise not only strives to promote women, but also endeavors to forge connections between them,” Kouki said in nominating Getoor for the Women in Science and Engineering Award last year. One of Kouki’s favorite memories from UCSC is of her 2018 commencement when Lise Getoor hooded Kouki and two other female Ph.D. students, the first three Getoor had graduated since she came to UCSC in 2013.
“Lise is a world-renowned researcher in her field, but also she is a great Ph.D. advisor,” Kouki says. “She taught me how to do world-class research, how important it is to be hard-working and persistent. Lise always sets the right example for all her students and she will always be a role model for me.”
Getoor was also the one who first introduced Kouki to her current employer, RelationalAI, with whom Getoor was collaborating at the time.
Kouki has now been with RelationalAI for a year and a half, and is currently working on developing recommender systems for large e-commerce companies. She likes to see how the work she is doing is helping people to have a more customized experience than they would be able to have without the recommendation engine.
“What I really like is the fact that my work has a direct impact on the users and the way that they shop and find new interesting products,” she says.
For students who are interested in following her path into industry, Kouki has some simple, but important advice: “Work hard, have your work and achievements speak for you. Be persistent, and do not give up until you get what you really love.”