Roomi

Inspiration

Roomi was inspired by a problem that every college student has, which is living with roommates. With forgotten chores, to awkward interactions. Many times small issues will spiral into large problems. I noticed that many conflicts between roommates comes from a lack of communication and understanding, so I wanted to build a tool that keeps everyone accountable, helping roommates stay organized, sharing responsibilities, and resolving conflicts with grace.

What it does

Roomi is an all in one AI powered roommate assistant that helps students. Find the most compatible roommates by matching profiles. Mediate conflicts with AI that helps come up with thoughtful messages, turning a tense message into a respectful one.

How we built it

The frontend was built using Next.js + React for a fast and responsive user interface. The backend was built using Next.js API routes running in the Node.js runtime The database is a Supabase for storing accounts, roommate groups, members, chores, and bills. Supabase is also used for account login and authentication. Used Google's Gemini 1.5 flash model via the @google/generative-ai SDK. This was used for finding matches, generating schedules, and helping mediate tensions between roommates. Along with a round robin scheduler that ensures chores rotate fairly between every roommate.

Challenges we ran into

I had issues with Gemini returning non-JSON responses, so I built a tolerant JSON parser and deterministic fallbacks. The Gemini SDK also only works in Node so each API route needed to be explicit. I ran into some challenges in designing a scheduling algorithm that is fair for all group members. Integrating the project with Gemini and setting up Supabase was challenging due to never using those tools.

Accomplishments that we're proud of

I was able to build out a fully functional MVP in under 24 hours using tools that I haven't used before like Supabase, and Google Gemini. Delivered a local impact by targeting one of the most relatable problems for students, which is roommate stress. Built and end to end full stack AI powered application.

What we learned

Managing the full stack end to end: database design, API routing, AI integration, and frontend UX in a very short time. I learned how to integrate Gemini AI safely with fallbacks. I also learned how to build and structure Supabase schemas for multi-user group data with relationships and real time syncing. Ensuring that if the AI failed the app would still be effective at producing deterministic results.

What's next for Roomi

The next steps for Roomi is expanding AI mediation to handle multi-message conversations. Creating a chat feature for potential roommates to learn more about each other, Implementing real time notifications. Expanding AI powered scheduler to feature more custom scheduling. Deploy the app for University students.

Built With

• gemini
• github
• next.js
• node.js
• react
• supabase
• tailwindcss
• typescript
• vscode