The Georgia Institute of Technology, Atlanta, GA Ph.D., School of Mathematics • Dissertation: Exponential Decay of Resolvents of Banded Infinite Matrices and Asymptotics of Solutions of Linear Difference Equations • Adviser: Professor Evans M. Harrell II • Area of Study: Differential and Difference Equations, Mathematical Physics, Special Functions M.S. in Applied Mathematics, School of Mathematics.
The University of Alabama in Birmingham, Birmingham, AL B.S., Department of Mathematics.
Coursera Classes Completed • Machine Learning, Instructor Andrew Ng, July 2012 • Computing for Data Analysis, Instructor Roger Peng, April 2013

Data Scientist, Macy’s Systems and Technology, Johns Creek, GA, February 2016 -Present
• Formulation of data science role to fit with company strategy and ”skateboard f irst” agile development.
• Formulatedwaystoquantify ”poor user experience” for Macys.com andBloomingdales.com user population so predictive models could be used to perform root cause analysis using graphs of connections between application JVM pools, number of transactions, and response times.
• Anomaly detection algorithms used to flag mis–behaving servers using CPU utilization, memory, etc. K-nearest neighbors, Median Filter, and standard deviation detectors were evaluated and put into production systems. Evaluated resulting system for false positives with management team.
• Worked with junior team members to build a tagged dataset of anomalies for use in evaluating models. Incorporated the tag information into the dashboard for review and further action by stakeholders.
• Tensor Decomposition of multidimensional data – application, server, item (CPU utilization, etc).
• Code and model reviews of junior team member’s work along with mentorship and guidance in work-related matters.
• Formulated plans to integrate disparate datasets from server logs and other sources which were collected at different time intervals. Formulated a plan to move all data collection to seven minute intervals in order to avoid timing and imputation problems. Rejected several proposed models to impute data from five minute to one minute intervals.

Data Scientist, Nexidia, Atlanta, GA, January 2015 -January 2016
• Used ensemble classification algorithms such as Random Forests to improve and rework existing churn prediction model for a major cable operator. Feature selection and algorithm parameters were selected using recursive feature elimination and searching a grid of parameters (coarse, then fine) using cross–validation. Loaded and extracted evaluation data provided by the customer. Prepared reports evaluating model against out-of-sample churn experience.
• Troubleshoot and improve existing production churn model in R. Wrote R code to train new version of the random forest model and evaluate current model. Evaluated multiprocessing infrastructure in R versus python. Rewrote production prediction process in Python using scikit–learn and pandas. Integrated new model into the existing production system.
• wrote a Cython wrapper around Nexidia C++ libraries used to read tens of thousands of transcript files along with metadata.
• Wrote feature extraction code to reproduce a SVM model from a published in order to classify zones (disclaimer, etc) in emails.

Senior Quantitative Analyst, Fiserv., Norcross, GA, March 2011 -December 2014
• Investigated auto loan prepayment modeling and formulated proposal based on Fed research papers for adding auto loan prepayments to prepayment model. Obtained prepayment history for over forty auto loan deals from Bloomberg for various manufacturers. Completed prototype code in R/Rcpp and wrote production code in C++ and Managed C++.
• Implemented bank balance sheet forecasts using Call Report data for proposed DFAST consulting business. Used R to investigate various methods for model selection and out-of-sample performance using the AICc and PRESS statistic. Implemented ridge regression method in C++.
• Wrote a proposal to add seasonality and trend decomposition to the existing time series analysis tool.
• Wrote a proposal to address prepayment model changes including fitting seasoning, burnout, and seasonality effects as well as a new two factor (home prices and interest rates) option-theoretic valuation methods using partial differential equations to incorporate default and prepayment in the boundary conditions.
• Evaluated the economic capital model. Wrote technical paper outlining the model and incorporated changes into the C++ source code. Wrote unit tests to assure that changes to the source only incorporate known changes. Reimplemented in R to check the model source code. Bivariate normal probability calculations being a key part of the model, known algorithms were investigated in order to achieve the greatest speed while preserving double precision.
• Imlemented the Hull-White model in C++ and used QuantLib classes to bootstrap yield curves for use with fixed-income instruments with embedded interest rate options.
• Implemented linear regression model classes and model diagnostics in C++ using Armadillo for use in automated model building in various software products as well as in DFAST stress testing.

Analyst (Contractor), Credit Suisse, New York, NY, November 2009 -March 2011

Quantitative Developer, Vicis Capital LLC, New York, NY, June 2007 -October 2009
• Wrote SQL stored procedures and C++ programs to create multithreaded programs for scenario analysis of convertible bonds. Wrote C# programs to automate downloading data from Bloomberg and populating into the database. Wrote C# front end to facilitate setting up new convertible bonds. Produced various charts and analyses in R. Interfaced Excel with Bloomberg API. Supervised end of day pricing for over $4 billion in securities.
• Supervised work implementing and evaluating the finite element method for the Black-Scholes partial differential equation in ANSI C++ which was used to value and calculate Greeks for warrants book. Reviewed C++ code and models for pricing equity options with an eye to performance enhancements and reliability.
• Wrote reposting tool in Excel/VBA to streamline and automate risk report generation for risk committee and investors. Streamlined scenario analysis generation for US and non-US convertible bonds.
• Worked with convertible bond desks in Hong Kong, London, and New York on traders blotter in Excel/VBA which was interfaced via Web Services with associated C++ .NET classes in the .NET calculation server and Bloomberg. Automated the end of day convertible bond analytics generation.

• C++ 7 years
• R 6 years
• Python 5 years