Austin, Texas, United States
Performance Modeling
Cross Barrels Desk
- Built tool in C++ for comparing multiple timeseries of depth level limit order book data - Created live dashboard to monitor and visualize core metrics, top level differences, and abnormal discrepancies - Applied data science techniques to identity core heuristics that signaled discrepancies between two datasets - Implemented parallelized, space optimized version of a Dynamic Time Warping based comparison algorithm in CUDA
- Developed multi-factor LS model for predicting temporal anomalies: considered rebalancing flows, dealer delta hedging,and time series analysis to build IN strategy - Validated new signals through feature-tailored statistical testing: utilized HAC standard errors due to autocorrelation and heteroskedasticity in time-series data, used difference-in-difference regression to control for seasonal effects - Dynamically set portfolio weights and leverage in real time based on signal strength, exponentially weighted covariance, and market volatility - Incorporated transaction costs and slippage in solving multi-period optimization problems - Worked with Polygon WebSocket, IB RESTful, and Bloomberg APIs to manage live data and trading
- Designed and programmed equity back-testing platform, modeling slippage and transaction costs - Created strategy for trading International Equities based on inter-market and intra-market relationships - Backtest Sharpe of 5.3, live Sharpe of 3.4, annualized returns 30%+ amongst weak international performance