Production ML systems, competitive predictors, and real-time AI applications.
World Cup Outcome Predictor
Probabilistic match and tournament predictor using historical international match data, calibrated team strength ratings, and Monte Carlo simulation. Processes 20+ years of match data to generate win probabilities and tournament forecasts.
10,000+ simulations per tournament • 2nd place (2 years)
💡 Built a system that consistently outperforms bookmakers—predictive accuracy translates to real value.
One-Shot Face Recognition System
One-shot face recognition pipeline using pretrained deep learning models. Recognizes identity from a single labeled image in live camera feeds with real-time inference.
Single image training • <100ms inference
💡 Real-time recognition from one example—minimal data requirements, maximum deployment flexibility.
Financial News Analyzer
Processes 1,000+ financial articles per day with sub-second semantic retrieval using embeddings and similarity search. Built for rapid information discovery in high-volume financial news streams.
1,000+ articles/day • <1s retrieval time
💡 Production-grade semantic search that scales—fast retrieval matters when markets move in milliseconds.
Every project here started with a real problem and ended with working code. I focus on building systems that deliver measurable impact—whether that's prediction accuracy, inference speed, or scale.
All code is open source on GitHub. Feel free to explore, fork, or reach out with questions.