Inspiration
Our school started a "Free Food Coalition" on GroupMe and it started growing exponentially to the point that it became unmanageable. Obviously we had to intervene.
What it does
freefood.club is meant to completely replace all the free food chats. It helps users find and post listings about free food (just as they would on a GroupMe) with the added benefit of being able to see the listings in an organized manner. Think Craigslist for free food.
In addition to manually posts about free food, this app also has a GroupMe chatbot component, where it uses machine learning to identify the legit food listings among the constant barrage of [mostly] irrelevant messages. This allows the app to easily "import" content from the countless free food chats that pop up everywhere.
How we built it
- We used React for the frontend to build a very minimalistic and clean progressive web app.
- We used Firebase for authentication, databases, storage, hosting, as well as cloud functions and even the brand new Firebase Extensions.
- We used Google Cloud Vision API for recognizing and extracting text from images/fliers posted on GroupMe.
- We used Google Cloud AutoML for both classification and entity extraction, which helped us confidently recognize meaningful posts.
Challenges we ran into
- Firebase and Google Cloud suffers from poor documentation as always.
- Collecting and cleaning data to train our ML models took way way longer than we expected.
- Getting a PWA to behave consistently across different devices was a challenge.
Accomplishments that we are proud of
freefood.club is a functional app that is already ready to use by everyone.
What we learned
- React is great for making lightweight yet powerful web apps.
- Natural language processing is even more difficult than images.
- Preparing data is 90% of the work in ML.
What's next for Free Food Club
A stable release with everything fixed and working as intended, so that we can spread our app across the nation.
Log in or sign up for Devpost to join the conversation.