Deep dives into quantitative finance, predictive modeling, and analytics. Where logic meets execution.
Engineered robust financial models focusing on non-Gaussian tail risk via GARCH volatility forecasting, Value-at-Risk (VaR), and Expected Shortfall (ES). Backtested Pairs Trading strategies utilizing Co-integration and Vector Autoregression (VAR) to exploit mean-reversion signals while optimizing the Information Ratio.
Developed a cutting-edge deep learning architecture to approximate complex derivatives pricing functions. By deploying Neural Networks as "Deep Surrogates," this approach drastically reduced computational latency compared to traditional Monte Carlo simulation baselines, ensuring high-frequency viability.
Prototyped LLM-assisted, scene-aware multi-modal CTR features. Executed rigorous model sweeps across XGBoost and embedding architectures. Designed and analyzed online A/B experiments for ranking strategies with strict temporal alignment, successfully delivering a 12% CVR lift.
Formulated atmospheric pollution source identification as an ill-conditioned inverse problem. Applied linear algebra-based estimation methods and MATLAB simulations to localize emission sources, stress-testing the algorithmic stability under extreme measurement noise.