Heirloom AI

Timeline of a Family's shared history
Transcribed information is shown upon inspection!
Upload and transcription for new data

Inspiration for Heirloom 🌟

More than ever, it's difficult to be close to your ancestors. We are a team of immigrants and foreigners and we understand the pain of growing up in a culture that is so different than who came before us. Because of this, we had to adapt to this new culture, which in return of assimilating and integrating within a foreign environment, lost much of our culture and identity. Through the various types of documents that those before us had passed on, we wanted to connect our past with our present by preserving and understanding what has been given to us as precious keepsakes.

What Heirloom does 💪

Heirloom is a platform that lets you keep, share and easily read family documents, or heirlooms. We are able to take heavily degraded documents, and convert them to text, using NLP to patch missing information. Then, you can view and share the documents in your immediate family, or share to the world so that your documents are never lost in time.

How we built Heirloom 🔨

Our tech stack was NextJS and TypeScript. We used various component libraries, most prominent being DaisyUI. We also primarily used Together.AI & HuggingFace for our inference and transcription capabilities. It was built using a fractal folder design pattern, along with a singleton instance pattern.

Challenges we ran into ❌

OCR is difficult! We had issues with it, but we came up with novel ways of patching and augmenting data to perform better on our model. We also had difficulty connecting the front-end with the back-end, in addition to not being able to fully complete our transcription. model.

Accomplishments that we're proud of 🏆

Being able to understand and adapt to new technologies that were foreign to us. Almost all of our team was unfamiliar with the technology found in our current project.