Every year, on Valentine’s Day eve, thousands of Cornellians log into their emails, awaiting a message from Perfect Match revealing their potentially perfect Cornell love.
Perfect Match is a student-developed matchmaking service for Cornell community members that matches participants with potential romantic partners for Valentine’s Day using a machine learning algorithm. As of Feb. 10, over 3,600 Cornellians had already submitted their Perfect Match survey, according to its website.
The survey asks participants questions about their interests, academic pursuits, relationship preferences and dealbreakers, according to Elif Yilmaz ’27, Perfect Match’s treasurer and business team lead. The matchmaking algorithm then reviews survey responses, assigns compatibility scores across users and returns top matches.
Playing Cupid for Cornellians
Founded by Jamal Hashim ’22 in 2019, Perfect Match was originally developed to uplift spirits on campus and reduce isolation among Cornell students. Seven years later, Perfect Match still reaches and connects thousands of Cornell students, alumni and faculty members through its specialized survey.
The current Perfect Match team consists of nineteen undergraduate students, split between the engineering and business teams. The engineering team builds the website interface, tunes the matching algorithm and visualizes survey results. The business team conducts user research, reaches out to potential partners, manages Perfect Match’s image social media platforms and organizes events to increase engagement.
The team is tight-knit, Daniel Chuang ’26, engineering team lead and data analysis subteam head for Perfect Match, said, with each member contributing to the shared goal of uplifting and connecting Cornellians. For Yilmaz, Perfect Match’s romance-driven mission and start-up energy drew her in initially.
“I wanted to be in the startup environment where it's all about love,” Yilmaz said. “I'm a lover girl. That's honestly why I joined.”
Chuang initially joined to apply his interest in data analysis and visualizations to an engaging and widely viewed project.
“Something that I care a lot about is if I'm putting in a lot of time to make analyses, I want people to see it,” Chuang said.
A One-of-a-Kind Experience
Perfect Match offers a unique way for Cornellians to find partners and friends outside of more common dating services. The Perfect Match team cited how popular dating apps like Hinge and Tinder place extra emphasis on the photos that participants submit rather than personality and compatibility.
“On Hinge, you get to choose two or three prompts, and you see just answers to these three prompts, some of their lifestyle choices and their face … It doesn't really say much about someone's persona,” Yilmaz said. “I think one of our biggest differentiators is asking those very detailed questions, and also asking who you are looking for in a partner, and making matches based on that.”
Students might be more interested in filling out Perfect Match because it attracts a group of Cornellians that is “different than the typical dating app pool,” Chuang said. He added that Perfect Match typically gets higher participation from both females and males, which is uncommon for traditional dating apps, which are typically male-dominated.
Other college campus matchmaking services like Marriage Pact, which is an annual matchmaking survey that started at Stanford, also collect survey responses for research, but Chuang said that Perfect Match operates with the singular motive to connect Cornellians.
“We’re doing something just out of good will,” Chuang said. “It’s just to make Cornell a hopefully happier, funner place to be at.”
Crafting the Perfect Survey
The Perfect Match survey asks unconventional questions like “What is your Rice Purity Score?,” as well as Cornell-specific questions like “Biggest niche Cornell red flag?” and “Next year, I want to complete this task (from 161 Things Every Cornellian Should Do)...” To develop these questions, the entire Perfect Match team works together to brainstorm, starting months before Valentine’s Day.
The team draws upon their own experiences to draft the survey, and prioritize asking “funny but relatable” questions, Yilmaz said. Chuang added that they think of the questions in terms of three perspectives: the algorithm, the visual analysis and the fun.
“I think [the survey questions] are quirky,” said Joyce Shen ’27, head of algorithms subteam for Perfect Match. “They’re fun. They tie into the community. It’s just something that's a little more close to the heart.”
The team meets and brainstorms question ideas during the first semester, hoping to generate some new questions from year to year to maintain engagement and interest.
This year, the survey includes new questions to cover topics like spoken languages and expected graduation year for data analysis and visualization purposes. The 2026 survey also includes more analytical questions regarding political stance.
“Previously, one of the pitfalls I saw was that we just said politics is just ‘are you left or are you right?’ and that's something that really doesn't capture the nuance,” Chuang said.
Struck By Cupid’s Algorithm
Once the questions are set, the algorithms subteam focuses on tuning the machine learning algorithm to correctly evaluate survey responses and output compatible matches.
Currently, the algorithm is heuristic, meaning that the model focuses on quickly and efficiently identifying recommendations based on a set of rules.
“[Our algorithm takes into consideration factors like,] do these two people select similar questions based on their preferences?,” Shen explained. “People will sometimes say they want somebody who's more similar to them or more different to them. But we take all these things into account.”
Each survey question is mapped to a specific part of the algorithm, and numbers are assigned to responses accordingly. Survey questions regarding preferences, dealbreakers and mutual crushes are weighted more heavily than the more niche, interest-based questions, Shen said.
Shen also explained that users’ compatibility for each question has a different weight attached, meaning that certain questions may carry more of an impact to the final compatibility scores than others.
“If people have love languages that are more compatible, that boosts your score by a certain number, versus if you know both people have the same dream date spot that might boost your score by a different number,” Shen said.
From the survey responses and assigned weights, the tool calculates compatibility scores between users and users are sent their top matches.
Currently, the team relies on a manual check to ensure that every survey question has the appropriate weight assigned. The team also recently implemented a validation system to ensure that most users received an appropriate number of final matches. This need for a multi-step validation process stems from the challenge of training the machine learning model without concrete feedback of the success rate of matches. Perfect Match sends a feedback form out after the results are released, but it is difficult to enforce responses, Shen said.
“It’s really difficult to get feedback after the algorithm comes out,” Shen said. “It’s very rare that people will actually tell us if they met up with their match and how that went.”
Shen emphasized that the only strategy users should use when filling out the survey is answering honestly.
“Be genuine,” Shen said. “The algorithm is not trying to trick you or anything. There’s not any hidden boost that will match you with someone else.”
Finding Your Perfect Match
For Cornellians interested in meeting their future partner, making a new set of friends or filling out a fun survey, the Perfect Match survey will remain open until noon on Feb. 12.
“If you want to make friends, if you want to find someone, if you want a situationship, we have that option too … There is absolutely no downside,” Yilmaz said. “Just fill it out and see what your matches are. Maybe you actually end up finding your perfect match.”
According to the Perfect Match website, matches are expected to be released by email on the evening of Feb. 12. Until then, the Perfect Match team encourages all Cornellians to submit their survey responses and try their luck at Cupid’s algorithm.
“Do it for the plot,” Shen said. “Who knows what might happen?”
Teju Vijay is a member of the Class of 2029 in the College of Arts and Sciences. She is a contributor for the News department and can be reached at tv95@cornell.edu.
