Shoshana (Bluma) Gelley Email & Phone Number

New York City Metropolitan Area

Managed a team creating marketing-mix models to determine optimal allocation of funds between marketing channels for four international markets, refreshing semi-annually. Mentored and coached direct reports and members of adjacent team in both hard and soft skills. Leveraged background in academic research to research and guide improvements to models.• Reduced modeling time from 2 months to 2 weeks through automation and efficient task allocation.• Updated model to Bayesian framework, increasing model flexibility and usefulness for investment planning.• Led team in producing superior work despite geographic distance and a 9.5-hour time difference.• Managed communications with partners, driving a cross-functional initiative that increased partners’ acceptance of model results by 50%.• Was assigned a direct report with no experience or relevant coursework. Through intensive training and coaching, turned him into a significant contributor within a few months and a star contributor in 1.5 years.• Led team in automating and improving the modeling process, including updating to a Bayesian framework.• Successfully led a cross-time-zone team to produce superior work despite geographic distance.• Worked closely with partner teams to share requirements, data, and results.• Leveraged background in academic research to research possible improvements to models.• Served as SME in both our model and Data Science for direct reports and members of adjacent team.

New York City Metropolitan Area

Held sole responsibility for producing marketing-mix models for 4 major international markets and presenting results to stakeholders in non-technical teams.• Delivered annual or semi-annual refreshes for each market while working virtually cross-geography.• Automated the modeling process, reducing the turnaround time for each refresh from five months to two.• Received Edward P. Gilligan Award for Innovation for modeling work. Presented to only ten teams yearly, this is Amex’s most prestigious award.

Created and refreshed marketing-mix models to provide multi-touch attribution allocation of card acquisitions between marketing channels. Supported the Brand marketing & strategy teams with ad-hoc analyses when needed, utilizing Big Data tools & statistical analysis & modeling; communicated results to a non-technical audience. • Built data ingestion and cleaning pipeline in R to integrate data in multiple formats from diverse sources, reducing data ingestion time from 2 weeks to 30 seconds.• Conducted in-depth study of the freelancer ecosystem and Amex merchant database to determine our ability to identify freelancers in cardmember base.• Analyzed survey response data to extract information about how cardmembers feel about Amex; built models to predict likelihood of cardmembers to recommend Amex to friends.• Created dashboard for cardmember and acquisition metrics that was presented to senior leadership.

Greater New York City Area

* Research interests: Data Mining, Social Computing/HCI, and Social Network Analysis. * Thesis topic: mining and analyzing data from online social networks and peer-production communities to understand user interaction, then building tools to support user decision-making. Advisor: Torsten Suel. Research Projects:Detecting Notability on Wikipedia December 2014 – presentUsing machine learning to determine whether new articles fit Wikipedia’s inclusion guidelines; funded by the Wikimedia Foundation.* Choosing and extracting features; analyzing and visualizing data for feature selection using R with Shiny.* Building machine learning predictors in Python with scikit-learn.* Interfacing with the Wikipedia community throughout the planning and implementation process to solicit and respond to comments, suggestions, and functionality requests.Wikipedia Articles for Creation (AfC) Process March – November 2014Long-distance collaboration with other researchers to study the impact of Wikipedia’s AfC process.* Used MySql to write and optimize queries over tables with tens of millions of rows and determine the impact of user characteristics on the success of their articles.Predicting Deletion on Wikipedia September 2011—April 2013Using machine learning to predict whether or not Wikipedia articles will be deleted.* Designed a process for collecting difficult-to-gather data using Python and web crawling.* Cleaned data and extracted features using natural language processing, web APIs, and text processing.* Built a classifier that predicted with high accuracy whether an article would be deleted.*Worked with NYU Proteus (NLP) group on temporal information extraction. 2010-2011

Haifa, Israel

* Collaborated with 5 other team members to determine how a site policy change impacted user engagement.* Used R, Python, and Unix scripting/awk/sed to build a data pipeline to transform raw data for analysis.* Analyzed Yahoo! Answers usage data using R, created exploratory graphs, and conducted statistical tests.* Presented results to ~30 scientists at Yahoo Labs; submitted paper for publication in a leading conference.* Completed internship remotely and collaborated on writing paper using effective communication and tools such as ShareLatex and git.

Basking Ridge, Nj

Summer intern in the Collaborative Applications Research department at Avaya Labs Research. Worked on massive online social network crawling and analysis.

New York, Ny

* Worked together with Dr. Haimonti Dutta of Columbia University to explore novel methods for predicting epileptic seizures. * Used Java and Weka's API to build and train SVM predictors on EEG data. * Concluded that the combination of SVMs and Granger Causality showed promise for seizure prediction.

Developed functionality for new area of website.

Created database for advertisers with built-in coded functionality.

Doctor Of Philosophy (Ph.D.), Computer Science

Ms, Computer Science

Activities and Societies: Undergraduate mentor

Bs, Computer Science, Gpa: 3.91

Activities and Societies: Secretary, President, Touro College ACM Branch; Student counselor; Peer tutor