Every college student knows the struggle of needing a temporary item (like a suit for an interview), and then realizing buying it brand new is a waste of money. Meanwhile, thousands of dollars of perfectly good gear sit idle in dorm rooms just down the hall. Existing marketplaces are clunky and full of friction, so we built Campus Current to create a hyper-local, trusted exchange natively designed for university communities.
What it does
Campus Current is an AI-powered peer-to-peer exchange built specifically for UNC. Instead of making users click through rigid search filters, our platform lets students type natural, conversational requests. When someone types a casual prompt like wanting to borrow an interview suit, our backend fires up the Gemini API to instantly translate that text into a clean, structured JSON contract. The app then cross-references the request against live inventory, validating budgets, sizes, and deposit caps by evaluating financial logic such as Total Cost = Rental Fee + Deposit. Security is locked down with verified @unc.edu domain access via Supabase Auth and end-to-end handoff workflows.
How we built it
We built a modular full-stack app optimized for fast hackathon delivery. The backend runs on Flask, SQLAlchemy, and Flask-Migrate, using SQLite locally and PostgreSQL via Supabase in production. Flask handles direct integration with the Gemini API to convert messy student text into reliable transactional data with fallback handlers. The frontend uses clean Jinja templates with vanilla JavaScript, HTML5, and CSS3, and the whole stack is containerized with Docker and backed by comprehensive pytest test suites.
Challenges we ran into
Getting Gemini to consistently output strict JSON structures from unpredictable student slang was tough, which forced us to build tight fallback validation layers. We had to make sure that the 2 users couldn't accidentally reserve the same item from a lender. The security aspect needed to be addressed as well. This is where the image confirmation system came into play, where both the lender and the borrower would have to take pictures of the item before and after the transaction to confirm that the object was given and returned in an approvable state.
Accomplishments that we're proud of
We successfully shipped a fully live, production-ready app on Render with secure UNC domain authentication. Turning conversational student text into a smart marketplace matching engine using Gemini felt like a massive win, and our test suite kept our reservation logic completely bulletproof under pressure. We also created a dual-use app in which the user can request to borrow/buy items and also services such as tutoring/room cleaning/sports mentoring.
What we learned
We learned the hard way why defensive AI integration matters: treating AI output as untrusted input with strict parsing layers prevents crashes when models return unexpected text. Splitting work cleanly across backend intelligence, database models, and UI templates also let our team move fast without stepping on each other's toes.
What's next for
We want to expand beyond UNC to support micro-economies at universities nationwide. We also want to include a user rating/ranking system. In this system, the borrowers and lenders would receive a rating from one another at the end of every exchange. This would in turn increase the "level" of one user, which would give them greater and greater benefits during future transactions (less money to lend on hold while borrowing, the ability to buy higher value items). We also wanted to integrate a "Recommended for you" section in which an algorithm would suggest items to a user based on their previous transactions.
Log in or sign up for Devpost to join the conversation.