Projects

From open data to healthcare indicators
A Databricks, Spark and AWS pipeline that organizes ANS public data and supports analysis in Metabase.
Python · Spark · AWS · SQL
Full study in Portuguese
Language models for financial text
Adapting BERT models for sentiment analysis in asset manager letters, with documented methodology and evaluation.
Python · NLP · Transformers · PEFT
Full study in Portuguese
Machine learning for fraud detection
A public-data study of transaction classification and the trade-off between detecting fraud and raising false alarms.
Python · Machine Learning · Model evaluation
Full study in Portuguese

Fine-Tuning Language Models for Sentiment Analysis in Financial Markets
Using Domain-Adaptive Pretraining and PEFT to train language models to identify market sentiment in asset managers' letters.
Data Science
Machine Learning for Credit Card Fraud Detection
Building a credit card fraud detection model and exploring the trade-off between precision and recall, dimensionality reduction, and techniques for imbalanced classes.
Data Science

Dual Momentum: A Systematic Investment Strategy for Selecting Winning Assets
Using backtesting and Python, this study examines a Dual Momentum strategy with Brazilian-listed assets: BOVA11, IVVB11, and B5P211.
Investing
An End-to-End Data Pipeline Using Brazil's ANS Data with Databricks, AWS, and Spark
A project using public health-insurance data on Databricks to explore distributed processing, data modeling, Apache Spark architecture, and Metabase BI.
Data Engineering