Looking Back at DSC x BAC x Corner’s Datathon 2025

Last year’s highlights. This year’s inspiration.

The semesterly datathon has become a cornerstone tradition for the NYU Data Science Club. More than just a competition, it is the perfect opportunity for students studying data science to sharpen their creative and technical skills and collaborate with other like-minded students across different disciplines to tackle problems grounded in real-world relevance.

Our 2025 Datathon sponsor

Our partner for the 2025 Datathon was Corner, a rising social mapping app aiming to redefine how GenZ users discover places in the world around them. Rivaling other review-based platforms like Yelp and Beli, Corner focuses on capturing the vibe of places, (who typically goes there and how different types of people experience them) using sentiment and demographic insights to help users discover spots that truly match their preferences, whether that’s a restaurant, a nightlife spot, or a hidden gem worth exploring.

It should come as no surprise then, that the two prompts for last year revolved on understanding and leveraging data to draw insights about our generation’s preferences and habits. Both challenges were real problems that Corner was actively working on, and participants were provided with real anonymized data from the company’s production systems.

The first challenge pushed students to develop meaningful ways to best segment Corner’s user base into consumer archetypes based not just on demographic features, but on taste and their interests in places. The second challenge (and it certainly was a challenge!) asked participants to create a RAG-based system that turns a user’s query into relevant results using semantic lexical matching that feels more human and personal. For example, instead of searching for “italian food” or “restaurant” on traditional search systems, participants were tasked to build ‘Vibe Search’ that relied on more natural prompts such as “where to find hot guys” or “dance-y bars that have disco balls.” With intentionally open-ended prompts, teams had the freedom to experiment with their solutions that they delivered with impressive originality and creativity.

Through our conversations with the participants, one of the most striking aspects of the datathon that we came to appreciate was the diversity of its participants. Students from a plethora of majors as well as schools came together to form teams, and it was exciting to see a data science major in CAS partnering with a math major as well as a Stern finance student, each bringing a unique lens to the same problem.

The first challenge pushed students to develop meaningful ways to best segment Corner’s user base into consumer archetypes based not just on demographic features, but on taste and their interests in places. The second challenge (and it certainly was a challenge!) asked participants to create a RAG-based system that turns a user’s query into relevant results using semantic lexical matching that feels more human and personal. For example, instead of searching for “italian food” or “restaurant” on traditional search systems, participants were tasked to build ‘Vibe Search’ that relied on more natural prompts such as “where to find hot guys” or “dance-y bars that have disco balls.” With intentionally open-ended prompts, teams had the freedom to experiment with their solutions that they delivered with impressive originality and creativity.

Through our conversations with the participants, one of the most striking aspects of the datathon that we came to appreciate was the diversity of its participants. Students from a plethora of majors as well as schools came together to form teams, and it was exciting to see a data science major in CAS partnering with a math major as well as a Stern finance student, each bringing a unique lens to the same problem.

Kicking off the Datathon

For many participants, the datathon was an opportunity to apply classroom knowledge and work towards a solution to a real-life issue. “My goal with the datathon was to get a feel of the field at NYU, and to tackle a project specifically,” Hugo Lee, a Computer Science major said. His teammate, Yuyang, echoed this sentiment stating that he was here to apply some of the skills he’d learned this year onto a realistic scenario. “I’m in the Business Analytics Club’s machine learning team and we’ve learned a lot that can be applied to a project like this. I want to engage with some real world implementations.”

Hugo and Yuyang’s team approached the second challenge by dividing their project into two components. They first encoded and embedded written reviews of geographic locations and places, using neural networks to predict which place best matched those reviews. Their front-end allowed users to input a description of a feeling or “vibe” that the user was looking for in a location, which was then encoded and fed into the neural network to be processed into recommendations.

But they didn’t stop there. Wanting to push boundaries and be more innovative, according to Hugo, they added a creative spin to the prompt. In addition to vibe prompting for locations, they wanted to extend their system to using song lyrics for the same purpose. With this, users could input a song whose melody, instrumentals, and lyrics would be analyzed by neural networks to infer a specific mood and translate it into location suggestions.

Participants hard at work tackling real-world data challenges

Across the room, other teams came up with entirely different approaches to the two problems. “Now, we are basically trying to ideate the process of clustering.” Jacob, a sophomore studying finance and data science, said regarding the work his team was doing. “We’ve already clustered places using an algorithm and are now going to cluster user information and reviews.” As straightforward as they may sound, the execution proved complex. Jacob’s teammate, Daniel, said that the main difficulty for their team was determining the bucket sizes for their clustering algorithm. “If there are too few categories, then all the user profiles would be too similar; but with too many categories we would have trouble comparing users and reviews at all.”

Moments like these captured the essence of the datathon. Beyond coding and modeling, finding a proper solution to the prompt required dedication, creativity, and critical thinking about trade-offs, edge cases, and the nuanced intricacies of real data, mirroring the kind of challenges faced in industry settings. Indeed, for many students, the datathon served as a glimpse into what working in the data science industry actually feels like. Jacob reflected on the event as a good learning experience: “Before this, I didn’t even know much about clustering algorithms, but now I can definitely see myself working on projects like this full-time.”

Participating is always worth it for the free food

Of course, no datathon would be complete without its lighter moments, and last year delivered plenty. From meme and trivia contests to creative challenges like building personalized lists on Corner, participants found ways to have fun alongside the competition. And, as always, there was the universal motivator: free food.

And finally, after a weekend of innovation and hard work, one winning team was selected for each challenge. Corner’s founder and CTO, Jake Xia, stepped in as our judge, reviewing each project and selecting the ones that stood out for both technical depth and creativity. The winners were:

• **Easy Challenge: **Cornerstones — Andy Li, Taichung Wu & Travis Chew
• Hard Challenge: The Unfortunate Vectors — Tomas Gutierrez, Robin Chen & Yarden Morad

As we reflect on the success of the 2025 Datathon, it’s clear that the event continues to grow, not just in size, but in ambition and impact. This year, we’re excited to partner with Pulse Foundry AI to bring Datathon 2026, sponsored by Vercel and v0. Participants can look forward to prizes and recognition for top teams, employment opportunities at startups, connections to Google and other leading companies, as well as exclusive access to sponsors and industry networks.

Tune in for Datathon 2026!

With the event just around the corner, taking place April 10–12, 2026, we’re excited to see what new ideas, innovations, and collaborations will emerge. Whether you’re a returning participant or considering joining for the first time, one thing is certain: the datathon remains one of the best opportunities at NYU to build, learn, and create something meaningful.

We hope to see you there!

Article written by Franklin Dong, Ananya Mittal, and Grace Wang.