Quantitative
Breakout Predictor
A market-data pipeline that finds structure in the noise. Historical price data becomes technical features, then an XGBoost model scores potential breakout setups.
Inside the build
Built a historical price processor for stock and cryptocurrency OHLCV data. Vectorized transformations prepare multiple timeframes for anomaly detection and signal generation.
The model evaluates breakout setups with a 1:3 risk-reward ratio across five years of historical backtesting. Reported results: 40% faster preprocessing, 22% fewer false-positive signals, and signal generation under 15ms.
Project results from Yash’s résumé; historical backtests are not live trading results.
