Forrest Williams

Data analyst, election nerd, and bike-lane enthusiast

Skills

Python: pandas, numpy, requests, web scraping, data analysis
SQL: dbt, ETL/ELT, BigQuery
Dashboards: Hex, Looker Studio, Tableau, Streamlit
Visualization: matplotlib, seaborn, Plotly, Bokeh, Datawrapper
Machine Learning: scikit-learn (regression, classification)
Geospatial: GeoPandas, QGIS
Campaign Data: NGPVAN, Mobilize, Civis, Parsons
Collaboration: Git, Asana, Google Workspace, Jupyter

Hi, I’m Forrest, a Data Analyst based in Queens, NY.

I started my career in non-profit immigrant advocacy, then after several years with the Census Bureau, my path led me to Arizona where I was Deputy Data Director for the AZ Democrats/Harris for President campaign. I put this site together to show off some of my projects, tools, and skills I’ve learned along the way. (And I’ll probably post some personal tidbits, like my half year in Uruguay and Cuba following the 2024 cycle.)

Scroll down to see some (mostly) interactive snippets of my work. And click through any project to explore the data, code, and process behind it.

Recent Posts

Abundance NY City Council Endorsements Map

I made this for Abundance NY ahead of the NYC primaries to share a fun infograph of councilor endorsements, especially for down-ballot races where people may not know the candidates as well. This was mostly an excuse to use Datawrapper, since it is so universally beloved (for good reason!).

This was a quick map to build, but I did have to simplify geometries with GeoPandas and convert the shapefile for use in Datawrapper. It also got a lot of views!

Click through to see the full process.

Visit Full Post

Natural Language Processing and Campaign Messaging Strategy

I used topic modeling (NMF) to analyze social media posts from Marie Gluesenkamp-Perez (WA-03) and Chris Deluzio (PA-17), based on their 2022 campaigns.

This Bokeh scatterplot clusters MGP's tweets by topic to highlight patterns in messaging strategy.

Hover over points to read individual tweets within each topic.

At the end of the project, I compared topic frequencies between the two candidates using a Twitter-trained GloVe model.

This chart (also Bokeh) compares how often each candidate tweeted about key themes to highlight overlaps and differences in their messaging.

Visit Full Project Page

Maricopa Precinct v Closest Votecenter Dashboard

This interactive dashboard plots each voter against the nearest of 256 vote centers, exposing how precinct‐level aggregation misallocates resources in an open votecenter system. It supported the Voter Protection team in directing limited volunteer and legal resources to the centers where they were needed most in 2024 General Election.

Click through to learn more about the process and purpose of making this dashboard.