3D visual memory layer for AI chats that summarizes relevant past conversations using Hybrid search (keyword + semantic) and feeds relevant content to the LLM via Custom MCP so users never start a chat from scratch again.

Full-stack AI application that lets users query their own documents in natural language using a RAG pipeline to retrieve relevant context from a vector database and generate accurate, grounded responses using a local LLM.

A 'Spotify Wrapped' for typists: upload your MonkeyType data and get a personalized, animated breakdown of your typing, not a boring dashboard. Shipped to 500+ users.

Full-stack AutoML tool that makes machine learning accessible without code. Users can upload datasets, automatically preprocess data, train multiple models, evaluate performance, and deploy predictions. Features iterative hyperparameter tuning, model retraining, and real-time progress logs.

Flutter app that helps students and young professionals in Waterloo find housing through a Tinder-style swipe interface. Features hand-gesture swiping via camera-based detection, detailed property info, and listings sourced through a lightweight web scraper with preferences stored in Firebase.

Comprehensive ML model to visualize and predict greenhouse gas emissions using NASA APIs and Flask for web deployment. Implemented regression algorithms to analyze environmental data patterns and contribute to climate awareness initiatives.

Chrome extension providing real-time NBA and Champions League scores through API integration. Features live updates, team statistics, and personalized notifications for favorite teams and players.