As a part of the Search team, built and trained machine learning models for understanding the relations between web content and real world business operations. Developed experiments to measure the resulting improvements in search algorithm quality.Designed and implemented complex MapReduce pipelines for discovering patterns in business geo datasets. The output was used to enrich Google's Knowledge Graph and other data workflows.

Developed a Bayesian filter framework for real-time calibrations of arbitrary option price models. Integrated the filter into the firm's trading platform, including visualizations and rich user interface controls, which allowed a tenfold increase in the number of traded securities per user. Wrote a scripting layer for backtesting and optimizing the framework on large clusters.Created large high frequency datasets from historical equity market data, built an API around them for company-wide access. Produced interactive studies in Python and Matlab for trading strategy discovery, fast prototyping and performance analysis.

Created price and volatility forecasts for automated trading of equity derivatives and convertible bonds. Optimized portfolio trade sizes and order timing using robust covariance matrix estimators and Monte Carlo simulations.Developed scenario risk models for simulating market crashes; implemented the models into libraries for portfolio risk management in Perl, Tcl and C++. Used by risk managers, traders and back-office auditors.