I build data tools, from the dataframe backend to the browser.
I created and maintain Buckaroo (https://github.com/paddymul/buckaroo), an open source dataframe UI for Jupyter, marimo, and VSCode — summary stats and a histogram over every column, across pandas and polars. 683 stars. Scrolling millions of rows without eating the browser took a segment-merging row cache and parquet decoded client side.
Recently: Tallyman, a data science environment for coding agents. Named computations in a dependency DAG, results rendered live in a browser companion, every step keeping the prompt that made it. Demoed at PyData London 2026.
Before that: Flatfile, Attune, Tudor, and Bokeh/Wakari at Anaconda.
Looking for data infrastructure, developer tools, or agent tooling — where a hard backend problem has to surface as something a person can use.