Projects
Things I've built: recent AI agents that do useful work on a schedule, open-source R packages, and text analysis from my research years.
-
AI agent
pix
One visual analysis of the world, every day. A scheduled Claude Code agent picks a question, finds the data, runs the analysis in Python, and publishes a chart in a consistent house style.
-
AI agent
daily
A five-minute morning news brief, researched, fact-checked, and written by an agent. Small scripts gather headlines, market data, and viral posts; the model decides what matters.
-
Guide
ai-dev-stack
An opinionated record of what a data scientist's tech stack should be under three sets of constraints: enterprise Databricks, an early-stage startup, and open-source freelancing. Every pick traces back to a shared scoring rubric, so you can disagree with a score instead of a vibe.
-
R package
rtweet
An R client for collecting and analyzing Twitter data, widely used in research and teaching. Now maintained by rOpenSci.
-
R package
tweetbotornot2
An interpretable Twitter bot classifier built on gradient-boosted trees. Needs one API call per account and shows which features drove each prediction.
-
Text analysis
resist_oped
In 2018 I compared the anonymous New York Times op-ed from inside the Trump White House with tweets from Cabinet members to see whose writing it most resembled. The author, revealed in 2020, was a DHS official who wasn't among the accounts compared.
Looking for academic work? My publications are on Google Scholar.