Description as a Tweet:

Coterie is a social media app that promotes in-person socialization. Based on shared interests, it smartly matches people using ML models. Using peripheral data such as location and bluetooth, we build a deep understanding of an individual’s interests beyond what is noticeable.

Inspiration:

Do you remember the time you were a freshman at university? Or you just moved across the country, and you now do not know anyone around you? At each stage of our lives, we find ourselves in unfamiliar settings, struggling to make new connections. Of course, eventually, we find “our” people, but what about the struggle that comes before it? Coterie wants to ease this process for you.

Coterie is an application that matches us with others of similar interests. For example, for freshman, it would help us find those who engage in the same extracurricular clubs and attend the same classes. For young professionals, as they start their journey into their careers, they want dynamic relationships in both their professional and personal lives. There are so many use-cases of Coterie that we cannot begin to list all of them. Truly, Coterie helps you find “your people.”

What it does:

Coterie offers a way to analyze and store interests of an individual. Over time, the platform learns the major interests of its users, connecting them more accurately to others based on their frequent location visits and similar interests.

How we built it:

First, we ideated--thinking back on some of the struggles we faced in our day-to-day lives. Once we identified our problem, we began by building an API specification that was shared between the frontend and backend teams. We integrated our projects after every major milestone (registration, profile, matching, etc).

Technologies we used:

  • HTML/CSS
  • Javascript
  • React
  • SQL
  • Python
  • Django
  • AI/Machine Learning

Challenges we ran into:

We found it difficult to create a visually appealing design with dynamic views. An app that servers as a social media app often needs attention to details, and creating a full-fledged networking app in 36 hours was a challenging feat.

Accomplishments we're proud of:

We're really proud of the matching algorithm we created using K-means clustering. We have over 50 dimensions of data plotted for each user efficiently, and we use the K-Nearest Neighbors to find recommended connections. It was also fun to create a scalable API in Django and deploy it to Digital Ocean.

What we've learned:

We learned that communication is key while working towards a project that involves multiple components. Keeping to a API specification is interesting because it provides both the front-end team and back-end team a good guide to build their respective software.

What's next:

We aim to enable social media integration by extracting a person’s interests through their social media posts and type of content shared/liked on their accounts. This would allow us to more closely gauge a person's true interests and improve our matching algorithm's accuracy.

Built with:

Backend: Digital Ocean, Google Places API (Cloud), Django, PostgreSQL, Facebook Faiss, fastText
Frontend: React Native with Expo Framework

Prizes we're going for:

  • Best Software Hack
  • Best Web Hack
  • Best AI/ML Hack
  • Best Venture Pitch
  • Best Beginner Software Hack
  • Best Mobile Hack
  • Best Use of Google Cloud

Team Members

Arnav Nidumolu
Maneeka Mishra
Atharva Kale
David Gibb

Table Number

Table 26