Developed an API using Azure Function and automated processes with Logic App, optimizing customer support operations.

Built a predictive model to assess the likelihood of moderation requests being accepted, increasing approval rates by 50%.

Enhanced the bot's response with detailed reports, guiding analysts on how to improve their moderation requests, which helped elevate the company's platform rating from "regular" to "good".

Learned to navigate Azure tools efficiently, building a scalable API and automation workflow despite initial unfamiliarity with the platform.

Designed models to detect claims eligible for subrogation, using textual and structured data to improve decision-making.

Addressed challenges such as limited data quality and high-class imbalance, ensuring that relevant cases were accurately identified.

Enabled the company to recover approximately R$3 million per year, preventing revenue loss from overlooked claims.

Strengthened expertise in data cleaning, feature engineering, and handling unbalanced datasets to improve model reliability.

Developed a decision tree-based model to segment policyholders based on different risk levels, helping to define pricing and marketing strategies.

Analyzed a complex mix of structured and external data sources, including socioeconomic indicators and previous claim behaviors.

Supported a 7% reduction in claim rates, providing a data-driven foundation for product adjustments.

Gained deeper understanding of decision tree interpretability and data integration from multiple sources.

Developed models for hit-ratio (conversion rate after a quote) and loss-ratio (expected claim rate), optimizing pricing strategies for auto insurance.

Created a real-time monitoring system for conversion anomalies, helping adjust pricing dynamically.

Increased low-risk customer conversions by 9% and reduced high-risk conversions by 13%, improving portfolio balance.

Tackled challenges in granular loss-ratio prediction, ensuring meaningful insights despite small data segments.