Skip to content

LGeoff31/uwdorm

Repository files navigation

🏠 UW Rez: Find Your Perfect Dorm at Waterloo

UW Rez is the University of Waterloo’s first AI-powered residence recommendation and review platform, designed to help freshmen pick the right place to live — with confidence.

Now powered by a custom fine-tuned LLaMA 3 model, UW Rez offers rich, human-like summaries of your top matched residence — giving you a personalized overview of what to expect.


🚀 Why UW Rez?

  • Hundreds of students have already used UW Rez to explore and compare dorms at Waterloo
  • 🧠 AI-driven recommendation engine built using transformer-based embeddings + cosine similarity
  • 🧬 Now features a custom fine-tuned LLaMA 3 model to describe matched residences in natural language
  • 🗺️ Interactive map view of all first-year residences with markers and ratings
  • 🛏️ We rank dorms based on room quality, location, price, and bathrooms

💻 Architecture

image

🤝 Contribute

Lived in a residence at UW?
Help future students by sharing your experience!
You can rate and review your dorm — every voice matters.


🧠 How It Works

We use a custom-built AI system to pair students with their ideal residence based on preferences they provide. The engine consists of:

  • Transformer-based text embeddings for preference matching
  • Cosine similarity to compare user input with scraped housing data
  • A fine-tuned LLaMA 3 model trained here that generates a natural-language summary of the best-matched residence, describing room styles, location benefits, and bathroom types

📍 Features

  • ✅ Residence review system
  • 🤖 Smart AI recommendations
  • 🗺️ Map of Waterloo residences
  • 🏅 Rankings by room, location, price, and bathrooms
  • ✍️ Personalized AI-written summaries of matched dorms

⭐ Support

Love the project? Consider giving it a ⭐ to let us know!
We’re always open to feedback — feel free to open an issue or pull request.


About

AI dorm recommender

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Contributors 2

  •  
  •