Bridging rigorous numerical mathematics with systematic quantitative trading and deep learning.
Major contributions to open-source quantitative finance and algorithmic trading ecosystems:
Python library for valuation and risk management of financial derivatives.
- CDSCurve Vectorization & Day Count Alignment
: Aligned day-count convention handling in
CDSCurve.dfwith underlying discount curves (fixing hardcoded ACT/365 bypass), implemented seamless vectorization forDatelists across discount factors and survival probabilities (times_from_dates), and re-enabled comprehensive regression suites. - Shifted Black–Derman–Toy Model
: Extended BDT short-rate tree dynamics to support negative interest rates and updated lattice-based derivative pricing under negative yield environments.
- Equity Cliquet Option Valuation
: Implemented closed-form analytical pricing formulas and Monte Carlo simulation for return cliquet options with local and global caps/floors, validated with regression unit tests.
- Tech Stack: Python, Quantitative Finance, Analytical Pricing, Monte Carlo, Short-Rate Trees, Credit Default Swaps (CDS)
Open-source quantitative trading and strategy research platform.
-
Multi-Timeframe Duration & Metric Annualization
: Aligned monthly timeframe duration (2,592,000 s), annualization scaling factors (
$N = 12.0$ for Sharpe, Sortino, volatility), and driving frequency hierarchy across Strategy API V2, validated with comprehensive test suites. -
Factor Warmup Alignment
: Standardized lookback warmup requirements across technical indicators (EMA, HMA, TEMA, etc.) with parameter-aware dynamic resolution and unit tests to ensure reliable backtesting data feeds and factor evaluation.
- Tech Stack: Python, Quantitative Backtesting, Strategy API V2, Alpha Factors, Technical Indicators, Pytest
Symmetric factor orthogonalization and cross-sectional risk neutralization framework.
- Implemented Lowdin symmetric orthogonalization using SVD to eliminate cross-factor collinearity while minimizing signal distortion.
- Formulated Frisch–Waugh–Lovell QR projection for dynamic sector and market risk exposure stripping.
- Developed collinearity diagnostics (VIF, condition number, ENF) and cross-sectional Rank IC evaluation.
- Tech Stack: Python, Numerical Linear Algebra, Matrix Factorization, SVD, QR Decomposition, Pytest
Convex regularized alpha combination and dynamic multi-factor allocation engine.
- Formulated convex regularized objective balancing IC maximization, factor portfolio risk, and
$L_1$ turnover penalties solved via accelerated proximal gradient (FISTA). - Implemented Ledoit–Wolf analytical covariance shrinkage and Higham nearest symmetric positive-definite (SPD) projection.
- Integrated online exponentiated gradient adaptive weighting and Meucci Effective Number of Alphas (ENF) risk attribution.
- Tech Stack: Python, Convex Optimization, FISTA, Covariance Shrinkage, Factor Allocation, Online Learning, Pytest
Neural network-based differential equation solver in PyTorch.
- Built neural network architectures to solve linear, non-linear, and coupled ODE systems based on the Lagaris et al. framework.
- Formulated trial solutions that analytically satisfy initial and boundary conditions.
- Utilized PyTorch automatic differentiation (
autograd) for gradient computation and optimization. - Tech Stack: Python, PyTorch, NumPy, Numerical Analysis