Compass
DeployingAI-powered course recommendation system for University of Waterloo students using LangChain and modern web technologies
Overview
Course selection at UWaterloo can be overwhelming with thousands of options across different faculties. Compass solves this by using a custom-built recommendation algorithm that combines locally hosted embedding models with cosine similarity calculations and personalized weighting factors.
The system uses Neo4j to model complex relationships between courses, prerequisites, and academic pathways, while keeping all AI processing local for privacy and performance. The result is highly accurate course recommendations tailored to individual student goals and academic history.
Preview
Tech Stack
PythonFastAPILangChainPyTorchNeo4jReactTypeScriptLocal Embeddings
Key Features
Custom recommendation algorithm using cosine similarity and weighted scoring
Neo4j graph database for complex course relationship modeling
Locally hosted embedding models for privacy and performance
Interactive React frontend with real-time course search