Inspiration
As PhD students, we know the struggle: reading academic papers feels like a chore. Three out of four PhD students cite finding time to read as their biggest challenge, and when the average researcher reads 240 papers a year at nearly an hour each, that's ten full days just to stay current. We were frustrated with existing research tools—endless lists, dense PDFs, and no engaging way to discover what's worth your time.
Then we asked: what if exploring research papers felt more like... Tinder? Dating apps made learning about people fun and intuitive through swipes and progressive disclosure. Why not apply that same energy to academic papers? Yuzu was born from this simple idea: what if research discovery could be fast, delightful, and dare we say it—fun? 🍋
What it does
Yuzu transforms academic research discovery into an intuitive swipe experience. Search any topic, and Yuzu fetches relevant papers from ArXiv, presenting them as beautiful cards with AI-powered summaries at three depth levels:
- Swipe left to pass on papers that aren't relevant
- Swipe right to dive deeper into progressively detailed AI summaries (powered by Gemini)
- Press space to "superlike" papers worth saving
Each paper gets a quick Level 1 overview, a detailed Level 2 breakdown, and a comprehensive Level 3 analysis—so you can choose your depth. Save papers to your collection, export BibTeX citations with one click, and celebrate discoveries with confetti animations. What used to take an hour now takes 20 minutes, without losing depth or accuracy.
How we built it
Tech Stack:
- Frontend: Next.js 14, TypeScript, Tailwind CSS, Framer Motion for smooth animations
- Backend: FastAPI (Python 3.11) handling ArXiv API integration and Gemini AI requests
- AI: Google Gemini for multi-level paper summaries
- Deployment: Dockerized microservices architecture with Docker Compose
We built a responsive swipe interface using react-swipeable for mobile touch gestures and keyboard controls for desktop. The backend fetches papers from ArXiv's API and generates three levels of summaries through Gemini, while the frontend delivers a polished UX with confetti celebrations (canvas-confetti), toast notifications, and citrus-inspired design.
Challenges we ran into
Version consistency across Docker containers was our biggest headache. Getting the frontend to communicate reliably with the backend, managing environment variables across development and production, and ensuring dependencies played nicely in containerized environments took significant debugging time.
Other hurdles included:
- ArXiv API rate limits requiring smart caching strategies
- Crafting effective Gemini prompts to generate three distinct summary levels that felt natural and informative
- Making the swipe interface feel responsive on both mobile and desktop
- Balancing detail vs. readability in AI-generated summaries
Accomplishments that we're proud of
We're most proud of the UX design. Thanks to Cursor and aggressive "vibe coding," we built a complete, polished interface in hackathon timeframe that genuinely feels delightful to use. The swipe mechanics are smooth, the confetti adds joy without being annoying, and the progressive disclosure of information feels natural.
We also shipped a fully Dockerized, production-ready application with clean separation between frontend and backend, comprehensive keyboard shortcuts, mobile touch support, and one-click BibTeX exports—all in one weekend.
What we learned
"Vibe coding" can accomplish a lot—especially when combined with domain expertise. As PhD students living the research paper struggle daily, we knew exactly what pain points to solve. This personal experience made our design decisions faster and more confident.
We also learned that modern AI tooling (Cursor, Gemini) can dramatically accelerate development, but domain knowledge is what turns code into solutions. Understanding research workflows meant we could build features that actually matter, not just technically impressive demos.
What's next for Yuzu
Expansion & Monetization:
- Support for more databases beyond ArXiv (PubMed, IEEE, Google Scholar, bioRxiv)
- User accounts with cloud sync for saved papers across devices
- Currency/subscription system to "buy swipes" with tiered plans for power users
- Loot boxes to earn mystery rewards and gamify the experience for even more enjoyment
- Citation management integration with Zotero and Mendeley
Smarter Discovery:
- Improved recommendation algorithm that learns researcher interests and suggests relevant papers
- Collaborative features for sharing paper collections with lab groups
- Email digests of new papers matching your research profile
- Uploading papers you've found yourself for that special Yuzu zest
Because the future of research doesn't wait—it starts the moment you open the app. 🍋
Log in or sign up for Devpost to join the conversation.