Mathematical Engineering and Artificial Intelligence student at ICAI (Universidad Pontificia Comillas, Madrid). I build quantitative systems end to end — data pipeline, model, evaluation, and a usable interface — across derivatives pricing and hedging, portfolio construction and risk, mathematical optimization, and applied machine learning.
I'm primarily targeting quantitative research and trading, and the same core skills — statistical modelling, backtesting, optimization, causal inference, and automated research pipelines — carry directly into data science and applied ML roles. Each project below was designed and built solo.
Independent, end-to-end projects designed, built, and shipped solo.
| Project | Description | Stack |
|---|---|---|
| Chess Rating Leaks & Blunder Coach | End-to-end pipeline over ~19k of my own chess.com moves: local Stockfish labelling, leakage-safe features, a calibrated LightGBM model (≈3× the base-rate PR-AUC, 3.4× lift in the riskiest decile), a difference-in-differences study of whether clock increment reduces blunders (balance table, placebo test, E-value), and a self-contained interactive trainer built from the 300 costliest mistakes. | Python LightGBM Causal Inference Stockfish pytest + CI |
| OLPS Benchmarks · Bachelor's thesis (in progress) | Re-evaluates classical Online Portfolio Selection algorithms (Li & Hoi) under conditions the original literature skips: a survivorship-bias-free ETF universe, explicit per-trade transaction costs, and walk-forward hyperparameter selection — quantifying the breakeven cost at which each algorithm's theoretical edge disappears. A regime-aware meta-learner (HMM regime detection + feature-conditioned expert weighting) is layered on top, benchmarked against a static combination baseline. Data pipeline and Buy-and-Hold in place; further benchmarks, cost model, and regime detector in progress. | Python Online Learning HMM Portfolio Optimization Time Series |
| Options Pricing Simulator | Prices a real SPY option three ways — Black-Scholes, Monte Carlo, and a CRR binomial tree — cross-checks them against each other, computes Greeks closed-form vs. finite-difference, and benchmarks the methods on speed and use case. | Python Black-Scholes Monte Carlo Binomial Trees Greeks |
| Portfolio Risk & Performance Tracker | Python CLI that pulls a configured portfolio's prices from Yahoo Finance and reports risk/performance vs. a benchmark: Sharpe, Sortino, Calmar, Information Ratio, max drawdown, historical & parametric VaR, and CVaR. | Python yfinance Risk Management Performance Analytics |
| Shift Optimization — Fundación Proyecto Persona | MILP shift-scheduling optimizer built for a nonprofit's summer camps: Pyomo with Gurobi/HiGHS, Excel as the source of truth, and a Streamlit interface so non-technical staff run re-optimizations themselves. | Python Pyomo Gurobi HiGHS Streamlit Optimization |
Alongside the solo work above, I've built academic and team projects in supervised learning, NLP (classification, entity extraction, alerting), computer vision with OpenCV, mathematical optimization with Pyomo and Gurobi, distributed computing with Dask, Spark, and AWS EMR, and deep learning with PyTorch. Some originated in private or university repositories; cleaned public versions are linked where available.
| Project | Description | Stack |
|---|---|---|
| Strength Training Optimizer | Pyomo model that allocates training volume across goals, constraints, and per-session limits — a constrained-allocation problem with the same structure as scheduling and resource allocation. | Python Jupyter Pyomo Optimization |
| Discrete Mathematics Coursework | Modular arithmetic, RSA cryptography, and graph-based GPS routing. | Python Algorithms Graph Theory Cryptography |
| Project | Description | Stack |
|---|---|---|
| Football Commentary NLP | End-to-end NLP pipeline that turns unstructured match reports into structured signals: outcome/sentiment prediction, named-entity recognition, optional OCR, and alert generation. | Python NLP NER Sentiment Analysis OCR |
| Student Performance Grade Prediction | Regression models predicting students' final grades from academic-performance features, with a focus on model comparison and feature analysis. | Python Jupyter Scikit-learn Pandas |
| Project | Description | Stack |
|---|---|---|
| Computer Vision Final Project | Real-time OpenCV system: a visual-password workflow plus a two-player Snake game controlled by colored markers with Kalman-filtered tracking. | Python OpenCV Kalman Filter |
| Clover Pit | Unity slot-machine game with weighted paylines, store upgrades, adaptive difficulty, and physics-based lever interaction. | Unity C# ShaderLab |
