Leading analytics on the transformation of the bank's deposits infrastructure

Focused on deepening the understanding of customers' financial goals, behaviors, and interactions across products, branches, and channels through machine learning.

• Developed archetypes for Chase's 60 million customers through unsupervised machine learning techniques that helped inform product strategy, better target with advisor lead lists, and better serve up relevant content/marketing• Leveraged ensemble learning techniques to identify primary bank customers• Built ETL processes that helped digital product owners better understand the performance and usage of Chase's digital features such as Autosave and Chase First Banking

The Marketing Sciences team enables tracking and reporting on the performance of a client's media spend and delivers data driven insights & recommendations to improve the performance of their media campaigns. The team specializes in Marketing Mix Modeling (MMM) which is the use of statistical analysis to estimate the impact and predict the future impact of various marketing tactics on the client's KPI. Some of my projects and responsibilities were the following:• Built predictive models using Markov Chain Monte Carlo (MCMC) and Generalized Linear Models (GLM) to measure effectiveness of advertising in different media channels and external econometric factors that influence the KPI (sales, volume, brand image, etc.) of client's business.• Developed gradient boosted trees models to identify metrics with the strongest relationship to digital marketing campaign efficiency. • Implemented optimizations to maximize the return of ad spend based on marketing mix model results and presented recommendations to help clients hit their business goals. These optimizations resulted in a 19% increase of brand lift and a 7% increase of revenue for the client.

• Designed and executed quantitative research focusing on audience behavior, search, and advertising effectiveness• Partnered with engineering and analytic planning teams to construct a flexible and configurable data ETL and QC process for modeling data sets. This process built in python reduced data set building time by 90% and can be easily used by non-technical analytic planners• Helped clients maximize their media investments by applying the full scope of marketing science tools in collaboration with the agency’s planning and buying teams• Consulted clients from transportation, consumer packaged goods (CPG), and financial services industries

Metis is an immersive program that focuses on enhancing skills used in the data science industry. Using Python as a foundation, we approach problems with a goal in mind using a variety of statistical modeling, machine learning, and data acquisition techniques. Over the course of this program I completed 5 data science projects involving cleaning, processing, and aggregating data. I also trained various supervised and unsupervised learning models. Below are some of my projects.Catching Fish with Neural Nets• Trained a Convolutional Neural Network (CNN) to classify fish species with an accuracy of 95%.• Utilized Keras and Tensorflow for analysis, image augmentation, and GPU parallelism.• The model created achieved a top 5% ranking on Kaggle.Natural Language Processing on Wikileaks• Scraped and collected 51,000 emails from Wikileaks using BeautifulSoup. • Analyzed the Podesta emails using unsupervised learning methods and latent dirichlet allocation (LDA) to extract topics. • Applied algorithms to create a network visualization of the emails using Gephi.Classifying Obese Counties• Implemented supervised learning models to classify counties based on obesity rates.• Conducted analysis and modeling with pandas, numpy, and scikit-learn• Created a visualization of the predictions using the Seaborn library and d3.js.Predicting IMDB Scores• Incorporated Pandas and BeautifulSoup python packages to retrieve and manipulate data for more than 17,000 movies. • Employed gradient boosted trees and linear regression to determine the quality of a movie before release.MTA Traffic Analysis• Identified the turnstiles and subway stations with the highest traffic at different times during the week and used those insights to recommend the placement of street teams.