Caden Sun Quaden2307

About Me

caden = {
    "school":    "University of Waterloo - B.Math (2025 to 2030)",
    "focus":     ["Machine Learning", "Quant Finance", "AI Agents"],
    "currently": "AI Intern @ ProCogia  ·  ML Engineer @ Wat Street",
    "research":  "Member of Jimmy Lin's Research Group @ Waterloo",
    "loves":     ["basketball", "squid game", "cats"],
    "motto":     "ZOOWEEMAMA! - Rowley",
}

I'm Caden, a Mathematics student at the University of Waterloo who loves turning messy data into things that actually work. Right now I'm building AI calling agents at ProCogia, forecasting market volatility with graph neural nets at Wat Street, and chasing a sub-10% error on my flight-price predictor. I'm also a member of Jimmy Lin's research group at Waterloo, where I work on information retrieval.

What I'm currently working on (click to expand)

Jimmy Lin's Research Group @ Waterloo - information-retrieval research with the Anserini / Pyserini toolkits.
AI Intern @ ProCogia - built a LangGraph-orchestrated calling-agent POC (~500 ms faster, a 3x gain over Amazon Connect) plus a RAG pipeline on AWS S3.
ML Engineer @ Wat Street (Waterloo's Quantitative Finance team) - wiring a Graph Attention Network into an LSTM in PyTorch to forecast volatility for 500 S&P stocks.
Software Developer @ WARG (past) - shipped client-side mission-export tooling for the IMACS-3.0 ground control station (React + QGC .waypoints).

Featured Projects

Flight Price Predictor

End-to-end flight-price prediction across 200+ routes and 55+ airports. A SQLite pipeline ingests ~4k offers/day (180K+ rows) to train an XGBoost model targeting sub-10% error.

AI Chess Bot

A full-stack chess AI: a PyTorch evaluator (768 to 256 to 128 to 1) served via a Flask API and React/TypeScript frontend, Dockerized and deployed on Vercel with sub-second moves.

Chest X-Ray Classifier

A deep-learning study comparing models for classifying chest X-rays as normal vs. abnormal.

World Cup Score Predictor

A model that predicts World Cup match scorelines from historical match data.

RankLLM

A Python package for reranking search results with LLMs (pointwise, pairwise, and listwise rerankers) — one of Professor Jimmy Lin's group repositories, which I work with as part of his research group.