Economist turned data scientist — bridging rigorous analysis with practical ML and business impact.

Based in Montevideo, Uruguay, I'm an economist who moved into data science through academic research, teaching, and now industry. I spent years doing causal inference research and teaching Macroeconomics before moving into industry, where I now build fraud prevention and analytics systems at Mercado Libre. I hold two master's degrees from Arizona State University and the University of Georgia, and a bachelor's from Universidad Católica del Uruguay. I apply ML and AI hands-on through a postgraduate specialization at UTEC + MIT, and I've taught university-level Macroeconomics, Calculus, Statistics, and Applied Data Science.

• Download CV

Projects

AI & Machine Learning

Presentation · GitHub

Used Car Search Chatbot with RAG

Car buyers struggle to describe what they want in search filters. This chatbot lets them search 38,531 used vehicle listings through natural conversation — asking for "an affordable family SUV under $15K" and getting relevant results, with follow-up questions understood in context. Built with LangChain, FAISS vector search, and GPT-4o-mini. Final project for the Deep Learning module (UTEC + MIT).

LangChain FAISS GPT-4o-mini RAG Python

Presentation · GitHub · Colab

RoastAI: Computer Vision for Coffee Roast Classification

Coffee roast classification is typically done by hand — slow, subjective, and hard to scale. This project automates it with a ResNet18 computer vision model achieving 85%+ accuracy, with Grad-CAM to explain which visual features drive each prediction. Paired with 60+ years of global production data to put roast trends in business context.

PyTorch ResNet18 Grad-CAM Python

Presentation

House Price Prediction in Uruguay

Real estate pricing in Uruguay is opaque — sellers and buyers have little data to anchor expectations. This project scrapes listings from Mercado Libre and El Gallito (2023–2025), builds a CatBoost model tuned with Optuna, and achieves ~25% MAPE for Montevideo. Interpretability analysis reveals which neighborhoods and features drive price the most.

CatBoost Optuna scikit-learn Python

Data Analysis

Presentation

Montevideo Moves (Spanish)

Traffic incidents cost lives and public resources, yet most policy decisions are made without data. This project analyzes government records to find when, where, and who is most at risk on Montevideo's streets — and proposes targeted interventions backed by the data.

Python UTEC+MIT

Student Performance

What actually predicts student success — study hours, family background, or something else? This project uses econometric modeling and ML to find the factors that matter most for GPA, comparing the predictive power of linear regression and decision trees.

Python Regression

Covid 2021

During the pandemic, decision-makers needed fast, clear answers from constantly changing data. This Tableau dashboard tracks confirmed cases, mortality rates, and regional trends — built on SQL-cleaned data to give a sharp, scannable view of how the crisis evolved.

Tableau SQL

Customer Exploration for Marketing Campaign

Not all customers are worth targeting equally. This analysis segments customers by behavior and response patterns to help a marketing campaign spend smarter — identifying which profiles are most likely to convert and why.

R EDA

Statistics & Econometrics

US Presidential Elections

Electoral maps tell a story, but only if you can see the patterns. This project builds interactive visualizations in R to reveal how geographic and demographic voting preferences shifted across US presidential elections — making complex trends readable at a glance.

R Visualization

Reproduction Rights

When laws change, do people's concerns follow? This project uses Google Trends data to track shifts in public interest around reproductive health topics, connecting spikes in search behavior to specific court rulings and policy announcements.

R API

Regression Discontinuity

Does turning 65 actually improve your health, or is it just correlation? This replication of Carpenter & Dobkin (2009) uses a regression discontinuity design to isolate the causal effect of Medicare eligibility — one of the cleanest natural experiments in health economics.

R Causal Inference

Difference in Differences

Do minimum wage increases destroy jobs? This replication of Card & Krueger (1994) — one of the most cited papers in labor economics — uses a difference-in-differences design to challenge conventional wisdom, comparing fast food employment across New Jersey and Pennsylvania.

R Causal Inference

Get In Touch

Open to remote Data Analyst and Data Scientist roles — particularly in fraud, fintech, and CX/marketing analytics. Also happy to connect about research, collaborations, or just a good data conversation.