Abby Williams, PhD

Data Scientist and Assistant Dean

Abby Williams

Work

Projects

Shipped Products

Meridian

Career intelligence platform trained on 75,000+ cross-sectional records using ensemble ML and SHAP explainability. Serves 5,000+ active users with a 30% reduction in advising cycle time.

K-meansXGBoostSHAPAWS
View Project

CV-to-Resume Conversion Tool

Translates 200+ academic credentials into industry-aligned skills. Adopted by 5,000+ students without requiring ongoing staff involvement.

NLPStreamlitAWS
View Project

PhD Internships Database

Searchable career intelligence database with automated update pipelines. 10,000+ annual users across disciplines.

StreamlitAWS
View Project

Sentiment-Augmented Persona Validation

NLP pipeline validating career personas through sentiment analysis on 3,000+ open-text survey responses and 60+ interview transcripts.

TextBlobVADERNLP
View Project

COVID-19 PhD Career Impact Study

Interrupted time series analysis measuring pandemic impact on doctoral career aspirations (n=329). Published in Journal of Experimental Political Science.

PythonTime SeriesSTATA
View Project

NSF/Mellon Career Pathways Dashboard

Integrated ML persona predictions with demographic and geospatial data. Informed $1.2M+ in program funding allocation decisions.

TableauArcGISGCP
View Project

Dissertation Research

"Designing PhD Career Personas Using Cluster Analysis" - K-means, XGBoost, Random Forest, SHAP explainability, Streamlit, AWS. Full replication package on Harvard Dataverse.

RIASEC Composite Feature Engineering

Psychometric composites from Holland Code survey data as primary feature inputs for career persona classification across 75,000+ records.

PythonpandasEFAPCA
View on GitHub

K-means Persona Clustering

Four validated PhD career personas using K-means clustering with silhouette scoring of 0.37 across 75,000+ cross-sectional survey records.

scikit-learnPythonK-means
View on GitHub

XGBoost Classification and Model Validation

3-class XGBoost classifier (F1=0.58, precision=0.60, recall=0.57) with Random Forest and logistic regression benchmarks and SHAP explainability.

XGBoostscikit-learnSHAP
View on GitHub

Research

Publications and Presentations

How to Thrive Beyond Academia

Williams, A. and Williams, F. (2024). Public Humanities, Cambridge University Press.

Through Their Own Eyes: COVID-19 and PhD Students

Haas, N., Gureghian, A., Jusino Diaz, C., and Williams, A. (2022). Journal of Experimental Political Science, 9(1).

Modern Language Association Annual Convention

Presenter - PhD career development and program scaling.

Chronicle of Higher Education

Panelist - PhD career pathways and alt-ac transitions (with Stacy Hartman).

Connect

Get in touch