Projects

Selected work across AI/ML, forecasting, data engineering, and full-stack.

Tennis Match Outcome Predictor

Predict the winner between any two tennis players — even from different eras.

Python, feature engineering, model evaluation

• Built an AI model that estimates match win probability using surface type, player form, weather context, and historical head-to-head results.
• Supports hypothetical comparisons (ex: peak Djokovic vs peak Federer on clay vs hard court).
• Engineered features like serve/return strength, unforced error rate, and performance by surface.

Full-stack C2C ticket marketplace focused on trust and verification.

Angular, Django, SQL, REST API

• Designed listing flow, browsing and search, and buyer/seller experience for campus event tickets.
• Implemented account system and listing validation to reduce scams.
• Optimized the UI for clarity so buyers can verify that tickets are real before committing.

Short-term stock direction forecasting using historical pricing patterns.

Python, pandas, scikit-learn, time series classification

• Cleaned and aligned historical OHLC data (open, high, low, close) into supervised learning samples.
• Engineered technical signals (momentum, recent volatility, moving averages) to improve directional signal.
• Generated predicted vs actual movement and measured whether the model beats baseline chance over recent windows.

Probability model for match outcomes in the English Premier League.

Python, pandas, scikit-learn, public football APIs

• Pulled match data and team stats, then built features around recent form, home/away splits, and goals for/against.
• Produced win/draw/loss probabilities for upcoming fixtures.
• Used this as a sandbox to compare sports prediction and financial prediction methods.