Research

@ ETH Zurich

  • Currently writing a chapter for the Handbook of Economic and Language about large language models and their applications in economics.
  • Partnered with the Swiss Government (SECO) to develop a recommender system for the public employment service (RAV) to increase the quality of job matches. Integrated into the system strategies to reduce potential biases and explanations of the machine learning model predictions.
  • Developing a taxonomy of explainable machine learning techniques and a field experiment to test how they affect people’s perception of automated decisions.
  • Supported the development of three different courses.

@ UCL

I worked with professor Stephen Hansen to build models and tools for the use of unstructured data in economic analysis. Some of our main projects include:

  • Developing NLP models to predict the presence of remote work on more than 500 million job postings. A manuscript is currently in preparation and the language model can be tested online here.
  • Characterizing firms within a trade network through the use of embedding models. I presented a poster for this project at the CESifo Venice Summer Institute 2022 and currently maintain a public repository with all the code for estimating these embeddings.
  • Demonstrating methods for algorithmic text analysis in economics for the Annual Review of Economics. Current draft.
  • Measuring the evolution of economic uncertainty with word embeddings estimated on high-frequency news.

I have also collaborated with professor Andrés Álvarez in projects related to measuring income inequality and social mobility in Colombia. Some of our projects include:

  • Using modern language models (e.g. GPT-3, BLOOM) to extract family relationships from Colombian historical records. I illustrate this in a blog post here.
  • Developing a demo web application for the Inter-American Development Bank in order to estimate and visualize the distributional effects of macroeconomic shocks to different sectors of the economy. You can find the demo here.
  • Developing a simple web application that allows people to locate their household within Colombia’s income distribution. You can find the demo here.

Master’s

As part of my master program in Data Science at the Barcelona School of Economics I wrote a short group thesis analyzing the political discourse on Twitter of politicians participating in the 2017 elections from Germany, France and the UK. We used a Structural Topic Model in order to understand the main drivers of the narratives used by political parties. You can find the final version of the document here.

Undergraduate research

During my undergraduate at Universidad de los Andes I worked with professor Adriana Camacho performing data analysis to better understand public policy issues related to poverty, health and education in Colombia. Some of our main projects included:

  • Developing econometric models with panel data to understand the causal effect of a health policy intervention in Bogotá. Working paper.
  • Studying the evolution of household poverty in Colombia through a longitudinal survey. Article in Spanish.