Candidate for Bachelors of Mathematics Co-op student at the University of Waterloo specializing in Applied Mathematics - Scientific Machine Learning and Statistics, with a minor in Computing.
I build machine learning systems for financial services, with an emphasis on rigorous evaluation, interpretability, and production constraints. My work focuses not just on whether models perform well in isolation, but whether they are reliable, understandable, and genuinely useful in real-world settings.
ML Engineer — Wat Street (University of Waterloo)
Modeling volatility contagion across S&P 500 assets using Temporal Graph Neural Networks, with performance benchmarked against econometric baselines such as GARCH and HAR-RV to evaluate statistical and predictive validity.
AI Engineer — Dazia Consulting Inc.
Designed and implemented a full-stack AI tutoring system powered by Gemini 2.5 Flash for financial certification education, expanding boilerplate to production RAG pipelines using ChromaDB and Docker that processed 1,500+ pages of content and reduced token usage by ~30% via optimized retrieval strategies.
Built Express.js REST services with CORS and rate-limiting to enforce freemium usage constraints and support production deployment.
- Features: RAG context Question Generator, Gemini 2.5 flash AI Tutor Bot, Context and Topic Summarizer Cheat Sheet Generator, Flashcards, Daily limit set
Co-Founder — Wanderers
Building an offline-first social discovery platform designed as a social catalyst rather than a destination.
- The Goal: Minimize time in-app; maximize real-world interaction.
- The Engineering: Prioritizing validation over polish and designing for user trust rather than engagement metrics.
| Project | Objective | Tech Stack |
|---|---|---|
| PlainCents https://github.com/Kapil-Iyer/PlainCents | Automated expense categorization and investment tracking dashboard using K-Means clustering. | Python, scikit-learn, SQLite, Pandas, NumPy, Matplotlib, yfinance, PowerBI |
| RiskFecta https://github.com/Kapil-Iyer/RiskFecta | Implementing portfolio risk analytics (Efficient Frontier, Sharpe Ratio) for retail investors. | Python, PyTorch, scikit-learn, SciPy, PostgreSQL, Plotly, Bloomberg Terminal, Tableau, Streamlit Cloud |
| VectorMate https://github.com/Kapil-Iyer/VectorMate | Neural network chess engine (1000+ ELO) using CNNs and Minimax search. | Python, PyTorch, Flask, Hugging Face, Render, Vercel |
Languages:
Machine Learning:
Data & Infrastructure:
Web & Visualization:
- LinkedIn: https://www.linkedin.com/in/kapiliyer29
- Email: k22iyer@uwaterloo.ca
- Portfolio: https://kapil-iyer-portfolio.vercel.app
- Resume: https://drive.google.com/file/d/1E28OhEEVwSsBy0Xg-PxqwkxtBhjb3nf5/view?usp=sharing