@everyone

Inspiration

We were super inspired by the imaginative way Google Recap and Spotify Wrapped transform everyday data into captivating, personalized narratives. We thought: why not channel that creative insight into Discord, something we use on the daily?

What it does

@everyone takes raw JSON data from your Discord messages and transforms it into a rich assortment of insights. It parses your message history to generate a variety of custom, engaging metrics—including identifying your favorite topic through a novel ML pipeline. Using these metrics, we build a high-dimensional embedding for each user, which is then projected into a 3D semantic space where every user is represented as a point. This visualization lets you easily see your closest connections based on the similarity of your conversations. Additionally, we present these cool analytics alongside custom commentary tailored to your unique metrics, offering a personalized and engaging experience. Please take a look at our documentation in the Google Slides!

How we built it

We built the front end using Next.js and Three.js and the back end using Flask. Our system stores user embeddings and metrics in a SQLite database, while leveraging powerful libraries like NumPy, scikit-learn, NLTK, and other ML frameworks and APIs to compute our custom analytics.

Challenges we ran into

The biggest challenge we ran into was making our web app look clean and slick. For a project like this, the UI is especially crucial and we made sure to focus heavily on that!

Accomplishments that we're proud of

We're super proud of being able to implement the core functionality of our project, which was the 3d embedding space. We're also super proud of the interesting and novel metrics that we were able to come up with.

What we learned

While most of us were already comfortable with web development and basic ML, this project gave us a fantastic opportunity to dive deeper into dimensionality reduction and language modeling. We came away with a much stronger grasp on these techniques and enjoyed every step of the learning process.

What's next for @everyone

The first thing on our bucket list is hosting the website live! Also, we really hope to be able to implement in-website support for creating the JSONs. If this all goes well, something else we would like to do is add even more metrics!