Gettin Roomy with WPI

Finding a roommate at WPI is scattered across flyers, Discord, Reddit, and group chats. Our app centralizes it all, using AI to match students by living habits so you find compatible roommates.

Comment

Inspiration

Roommate searching at WPI is currently a fragmented mess. Roommate requests on campus flyers, Reddit posts, Discord chatter, and more; there is no central hub for students to find compatible living partners. We built Getting Roomy with WPI to consolidate this experience into a single intelligent platform designed by students, for students.

What it does

Users onboard by providing their graduation year, WPI credentials, and a detailed biography with interests. Once submitted, the Gemini API performs semantic analysis to pre-rank potential matches based on personality and lifestyle compatibility.

Instead of scrolling through endless lists, users interact with a streamlined, Tinder-like interface to swipe on personalized suggestions. Upon completion, the app generates a Match Report, a structured table displaying the user's "Yes" selections, their associated AI compatibility scores, contact emails, and essential lifestyle details to facilitate the next step in their housing journey.

How we built it

We used React for a responsive frontend and Node.js/Express for the backend infrastructure. The server hosts custom endpoints that interface with the Gemini API, serving as the brain of our matching algorithm. We offload the semantic analysis to the backend to ensure secure API handling and streamline the data flow between user inputs and the generative model's ranking logic.

Challenges we ran into

  • Button rendering across different browser engines
  • Exceeding the Gemini API free tier, we found a workaround by modifying the server to call the API in batches of candidates, instead of iteratively for each candidate.

Accomplishments that we're proud of

We are particularly proud of our pre-matching algorithm. By utilizing Google’s Generative AI, we move beyond basic keyword filtering to actually understand the "vibe" and nuances of a student's bio, resulting in much higher quality match suggestions.

What we learned

Hands-on experience in full-stack architecture and the UX design for "matching/discovery" apps. More importantly, we learned how to optimize LLM prompts to return structured, rankable data

What's next for Getting Roomy with WPI

Conducting a beta test with real WPI students to gather real-world feedback. We may also expand the platform to a mobile experience for a more seamless swiping experience on-the-go.

Built With

Share this project:

Updates