São Paulo, São Paulo, Brazil
Data Scientist
• Optimized the main identification machine-learning model to substantially reduce false negatives while maintaining false-positive rates, improving detection recall by more than 10%.
• Led first-party fraud reduction efforts for a major retailer, designing detection strategies and reducing fraudulent transactions. • Spearheaded cash-flow optimization initiatives, streamlining payment/reconciliation processes. • Coordinated cross-functional teams (fraud, finance, product, engineering) to translate business requirements into deliverable analytical solutions. • Developed and deployed data-driven models and dashboards to monitor fraud trends, enabling faster investigation and decision-making. • Mentored junior Data Scientists
• First Party Fraud - Led the creation of cutting-edge Fraud Detection models, leveraging Machine Learning to enhance accuracy and minimize false positives • Behavioral Security - Drove measures to thwart scams and social engineering threats. • Income Prediction - Developed a predictive model for estimating new customers' income levels. • Mentored junior Data Scientists
• Data Science Platformization - Helped migrate the cloud services from Google Cloud to AWS • Mentored junior Data Scientists
• Chatbot with NLU - Helped automate parts of the support flow and built a Rasa NLU model to handle open-ended dialogue, extending the chat beyond multiple-choice responses. • Ops Ticket Categorization - Built two models: an unsupervised ML model to identify ticket categories, and a real-time classifier to route incoming tickets to the appropriate queues. • Spam Classifier - Developed a real-time AWS SageMaker model to flag incoming spam messages
• Geospatial model - Created a recommendation algorithm for B2B activation • Recsys model - Created the first real-time and in-app model for customization, ranking the most helpful products for each customer in the home page. • Model Governance - Helped structure documentation and best practices standards for Data Science