Meg Savel, Ph.D. Email & Phone Number

Arlington, Va, Us

Worldwide, Oo

• Supervised and advised three research fellows as the team lead for a trust in government project.• Leveraged NiFi, Python, and SQL to aggregate data from multiple government entities to drive mission critical national security decisions.•Created document searching functionality using dynamic regex construction in order to build a reusable tool to handle event matches.

Alexandria, Virginia, Us

• Created text classification models using applied transfer learning techniques with pre-trained word embeddings including BERT, XLM, RoBERTa, ELMo.• Developed unsupervised and semi-supervised topic models using Correlation Explanation (CorEx), Latent Dirchlet Allocation (gensim), dynamic topic modeling (BERTopic, huggingface, transformers).• Queried Elasticsearch’s rest api to retrieve data from LexisNexis for machine learning model development.• Wrote web scraping utilities for data collection using Python and Selenium.

Fairfax, Va, Us

• Developed an ETL pipeline using SQL to connect data from SQL Server and Oracle databases to load into R for predictive analytics• Used R to develop a model to identify cases where veterans continued to draw benefits after their likely death to discover misappropriations due to user error or postmortem identity theft• Created methodologies for KPIs measuring possible health care provider fraud• Developed and refactored an image recognition model in Python using a neural net to detect possible cases of trademark infringement• Collaborated with team of software developers to productionalize machine learning model for government RFP

Chicago, Il, Us

• Produced mixed methods analysis of candidate and voter behavior with respect to gender role expectations that draws heavily on using data to reinforce theoretical concepts of political behavior• Developed textual analysis program for analyzing content of candidate webpages, debate performances, and campaign advertisements for senate races in 2016 and 2018• Designed a conjoint experiment (N=3,028) to measure the effects of gender and family characteristics on vote choice as a function of candidate and voter attributes

Chicago, Il, Us

• Co-authored a published book chapter on the lack of a decline in trust in institutions among the American public using General Social Survey data from 1973-2016 (N=39,991) • Created and implemented an original survey experiment (N=1,000) to measure and model simultaneous attitudes about political candidates (under review at Political Science Research Methods)• Wrote Python scripts to scrape every congressional candidate’s campaign website (over 800) in the 2018 midterm election for an ongoing text analysis project

Chicago, Il, Us

• Developed an original curriculum for an introduction to statistics course and lectured 19 undergraduate students about the fundamentals of statistics and inference in the social sciences• Designed lab sessions and problem sets in R and Stata to apply statistical concepts to real world datasets

Phoenix, Az, Us

• Completed an 8-week data science bootcamp for PhDs transitioning to data science.• Created a Flask app tracking all stocks traded on the DOW, NASDAQ, and AMEX markets using bokeh, pandas, and requests libraries in Python.• Constructed a boardgame recommender by analyzing 3.4 GB of data from BoardGameGeek’s API. The project used scikit-learn to run KNN models to identify similar, highly ranked games when provided a game title.

Doctor Of Philosophy - Phd, Political Science

Master'S Degree, Political Science

Bachelor'S Degree, Political Science; Women'S Studies