dumpy

Inspiration

Our inspiration for dumpy stemmed from our senior design project which explores the use of facial recognition technology. Working on that project sparked our curiosity about how similar AI-driven techniques could enhance real-world experiences. We began thinking about how artificial intelligence could help people organize and share memories before finally landing on the idea of dumpy.

What it does

dumpy turns shared moments into lasting memories. Whether it’s a concert, vacation, or spontaneous night out, dumpy lets you and your friends instantly upload photos into a shared dump. From there, AI brings your memories to life by automatically adding captions and setting them to music in a custom slideshow that fits the dump's vibe. Every photo is securely stored in the cloud, so your memories are safe and organized for whenever you are ready to relive them.

How we built it

dumpy is powered by a modern, scalable stack designed for performance and collaboration. The backend is built with FastAPI, delivering a fast and reliable API layer connected to a Supabase database for authentication and data management.

The React Native frontend provides a smooth, cross-platform mobile experience, while Azure Blob Storage securely hosts all user-uploaded media. To bring memories to life, OpenAI generates intelligent photo captions, and MusicGen creates custom background tracks to turn shared moments into immersive story slideshows.

Challenges we ran into

One of our biggest challenges was dealing with vague errors and silent failures in our facial recognition system. Debugging these issues without clear feedback forced us to dig deep into the model’s behavior and fine-tune our implementation under the tight deadline. We also faced difficulty finding an affordable and accessible AI model for music generation, which led us to explore multiple tools and adapt our workflow on the fly to achieve the results we wanted.

On the frontend side, the lack of planning time made it tough to scaffold the UI quickly while maintaining clean, reusable code, especially as we balanced design, functionality, and integration with the backend.

And, of course, in true hackathon fashion, one of our final challenges was simply securing a spot with dual monitors so we could all work efficiently! Despite the chaos, every obstacle pushed us to adapt, collaborate, and deliver a final project each of us are proud of.

Accomplishments that we're proud of

One of our biggest accomplishments was maintaining a clear separation of concerns within our team. By organizing responsibilities and communicating effectively, we minimized friction, increased development velocity.

We’re also proud of how quickly we adapted to new tools and technologies. From learning Figma for design to diving into Azure (after primarily working with AWS), we pushed ourselves outside our comfort zones without sacrificing quality or momentum. Finally, we’re proud to have built a fully featured social media application that integrates multiple layers of AI functionality all within a seamless and user-friendly experience.

What we learned

Throughout building dumpy, we learned that even the biggest challenges become manageable when broken down into smaller, actionable steps. By tackling one piece at a time, we were able to make steady progress toward our larger vision. Setting clear goals and milestones throughout the hackathon helped us stay motivated and made each achievement feel rewarding, turning what could have been a stressful sprint into an exciting, collaborative experience.

We also discovered how crucial communication and accountability are in a fast-paced environment. Not only keeping up with your own tasks but staying aware of what teammates are working on is crucial in staying aligned and on a forward track.

What's next for dumpy

We intend to use the facial recognition aspect of dumpy as a building block for a future project of ours. FaceIT is another project our team is currently developing. FaceIT is a facial recognition attendance app intended to streamline the tedious task of collecting attendance in classes especially during large lectures. In addition to it being a stepping stone for our FaceIT project, we are also interested in polishing dumpy, opening up a beta, and generating interest in it on social media.

Built With

• azure
• fastapi
• gpt-4o
• musicgen
• pytorch
• react-native
• supabase