McLovin (University Dating Site) - 2025
Django, PyTorch, MongoDB, Pinecone
My friends and I were sitting in an empty classroom one night, drowning in energy drinks and complaining about the McGill dating scene. That’s when it hit us like a caffeinated lightning bolt - why not build our own dating site just for McGill students?!
We called it McLovin.
The cool part? We made it super smart using machine learning to match people up. No more awkward “what’s your sign?” conversations - our algorithms do the heavy lifting! We started using a bunch of cool algorithms like K-means clustering (basically sorts people into groups of similar vibes). Also implemented a CNN for facial (selfie) recognition to assign specific preferences to users.
Then, with all that metadata, we would compare users embeddings with cosine similarity, and playing with some weighted averages, we made our own matching score.
mclovin’s brainstorm notes (PDF)
It’s still not live (Feb 2025), but our goal is to have at least a functional site before valentines…
LitCode (Multiplayer Coding Game) - 2024
Flask, Next.js, MongoDB
This was a project for CodeJam 2024. We won Best UI/UX Design, which is neat because we spent a lot of effort on that.
Basically, this is a multiplayer coding battle arena where you can absolutely destroy your friends with your algorithmic skills! We inspired ourselves with LeetCode. We built this beast using Flask and made it super secure (because nobody likes hackers in their coding battles).
The coolest part? During the competition, we hosted our game in a LAN under McGill’s public server, which made it simple and affordable (legit free).
I made a linkedin post about it…
And checkout this sickass demo vid lowkey took us an hour just to get the coding part done right:
Magic Chalk (Virtual Whiteboard) - 2023
Numpy, OpenCV, TensorFlow
This was our project for CodeJam 2023. While we didn’t win, we built something we’re really proud of. We wanted to make a whiteboard where you could just wave your fingers in the air like a wizard 🧙♂️ and have your writing appear on screen.
The real magic happened when we trained our own model to understand mathematical symbols and convert them to LaTeX, which we then fed into Wolfram Alpha! We built the interface using Streamlit, making it super easy for users to draw and get instant mathematical computations.
The thumbnail for the demo vid does not appear, so here is a pic of my friend eating a donut after submiting our devpost.
CLICK HIM FOR THE DEMO!
