The solution to many contemporary problems, big or small, begins with the question of measurement. My research focuses broadly on measurement problems in accounting. These include the prediction of accounting estimates, the analysis of corporate narratives, and the estimation of plausibly causal effects. In my works, I pair traditional quantitative methods with novel techniques from machine learning, natural language processing, interpretable machine learning, and causal machine learning. I am driven by an intense curiosity and approach my work with a scientific mindset. I enjoy deep work and alternating between R and Python to harness the best of both worlds.

My research explores

• how machine learning can help managers provide decision-useful accounting estimates and reduce human bias,
• how natural language processing can be used to meter complex phenomena in firms’ capital market communications with financial analysts, and
• how transfer learning can advance the state-of-the-art in textual analysis in accounting research.

My tech stack includes

• pandas and the tidyverse for tabular data wrangling,
• ggplot2 for data visualization,
• rvest, Beautiful Soup, and Selenium for web scraping,
• PyMuPDF, gensim, spacy, and LLMs for NLP,
• prodigy for data annotation,
• tidymodels, sklearn, and DALEX for machine learning,
• pytorch and transformers for deep learning,
• Weights & Biases for experiment tracking,
• fixest and did for empirical modeling,
• DoubleML for causal machine learning,
• gradio for web applications,
• rmarkdown, xaringan, and Jupyter Notebooks for literate coding, and
• Git+ GitHub for version control.