Software

I build data science tools for economists and anyone who works with data. These projects aim to support common workflows and help with everyday tasks, mostly built with R.

joinery

joinery is a record linkage package for R. It matches messy, inconsistent records across datasets by breaking them into weighted tokens and scoring on shared informative terms, so rare identifiers count for more than boilerplate. It supports out-of-core matching with DuckDB, scaling to data that does not fit in memory.

tidyllm

tidyllm is an R package designed to access various large language model APIs, including Anthropic Claude, OpenAI, Google Gemini, Perplexity, Groq, Mistral, and local models via Ollama or OpenAI-compatible APIs. Built for simplicity and functionality, it helps you generate text, analyze media, and integrate model feedback into your data workflows with ease.

SIAB DuckDB

SIAB DuckDB reimplements the data preparation for the Sample of Integrated Labour Market Biographies (SIAB), based on the original Stata reference code by Wolfgang Dauth and Johann Eppelsheimer (2023), and targets the SIAB 7523 v2 covering 1975 to 2023. It has two arms that run the same eighteen steps: an R arm where dbplyr translates each step into SQL for DuckDB, and a Python arm where the steps are polars expressions and the tables live either in a DuckDB database or in a folder of Parquet files. Neither arm holds the whole dataset in memory, so the preparation runs on data larger than the available RAM, and both arms are tested column by column against the Stata reference on IAB test data.

Eddy's Papers - Semantic Econ Paper Search

Eddy's Papers is a semantic search engine for economics papers on RePEc used by researchers at over 60 different universities and research institutes world-wide. Traditional keyword search misses relevant research due to jargon, synonyms, and rigid matching. This tool uses LLM-powered semantic text embeddings to find economic papers by meaning, not just words—capturing hidden connections and context. See the full code here

Shoddyprinter

Shoddyprinter is a whimsical R tool that gives your beautiful ggplot2 charts the aesthetic of a poorly calibrated office printer. Think jagged lines, misaligned labels, and broken toner style — perfect for presentations with a sense of humor or commentary on data visualization.