David Welgus

--Maintained and enhanced recommender system for marketing within Capital One apps --Designed offline and online evaluation frameworks --Maintained spark-based feature engine serving over 100 customer features for marketing segmentation and machine learning models, responsible for deploying all code changes to production.

--Helped build a model that predicts whether open leads will convert. Techniques considered: random forest, logistic regression, and extreme gradient boosting. Selected extreme gradient boosting via cross validation. --Improved the above model by taking account of the temporal nature of the data. --Built an ETL pipeline that transforms raw sales tables into common finance metrics (e.g. churn, expansion, reactivation, new business, etc…)

--Building statistical models to analyze experimental and observational data​: techniques include linear regression, generalized linear models, instrumental variables regression, and random effects models --Merging, cleaning, and managing data --Using machine learning to build predictive models: techniques include random forests, decision trees, adaboost, regularized regression, principal component regression, and text analytics (bag of words models)