← Projects

Here’s a selection of my personal and academic projects — spanning full-stack apps, AI, and embedded systems.

Full-Stack Board Game Manager (ECSE321)
Developed a full-stack board game management app using Java (Spring Boot + JDBC) for the backend and Svelte + TypeScript for the frontend. Designed REST APIs, integrated an SQL RDBMS, and practiced Agile (Scrum) development with JUnit tests for reliability.

MNIST Neural Network Classifier — [GitHub]
Built a fully-connected neural network from scratch using Python, NumPy, and PyTorch, achieving 95% accuracy on MNIST. Tuned hyperparameters, used dropout regularization, and visualized gradient/loss updates with Matplotlib.

School Supply Management App (ECSE223) — [GitHub]
Created a JavaFX/FXML desktop application using MVC architecture and UML design. Implemented extensive JUnit + Cucumber/Gherkin tests with 90%+ coverage. Focused on modularity and clean UI experience.

AutoBot AI Chatbot — [Devpost]
Winner of McGill CodeJam 14 (Matador AI Challenge). Built an AI-powered chatbot for car dealerships using named entity recognition and aspect-based sentiment analysis. Designed custom recommendation logic and outperformed 10+ teams.

RAG + AgentKit Integration Research
Explored Retrieval-Augmented Generation (RAG) with OpenAI AgentKit for scalable, context-aware chatbot capabilities. Tested pgvector + HNSW retrieval methods, studied embedding variance, and implemented linear truncation to improve latency by up to 70%.

MQTT Telemetry System (McGill Rocket Team)
Configured MQTT communication between a ground station and Teensy 4.1 microcontroller via Mosquitto broker for real-time telemetry. Enhanced link reliability and latency through low-level protocol debugging.