I develop an LLM-based agent that reviews code changes and orchestrates multi-step workflows to enrich contextual information and resolve security vulnerabilities directly within production pipelines.

I compare large language models against traditional NLP and machine learning methods to classify legal texts into procedural stages, and deploy the resulting pipeline into production.

I built explainable machine learning models to predict student dropout, combining SHAP and LIME with open-source LLMs to translate predictions into plain-language explanations for university staff.

As part of my technical studies at IFPB, I co-developed a hardware device with embedded sensors to measure and alert on excessive noise levels in shared environments such as classrooms.

A WhatsApp/Telegram/CLI assistant that explains the Brazilian legislative process in plain language. It answers questions about bills, votes, amendments, and the budget, keeps conversational context across follow-ups, and adapts tone and reading level (including audio) for users with limited formal education.

A public web platform mapping parasport and adapted-sport initiatives across Brazil, with zero-reload navigation across country, state, and city levels. Built with Next.js, React, and TypeScript, using d3-geo with IBGE territorial data for interactive SVG maps.

A cost-efficient system for extracting structured data from text-based PDFs, combining local text extraction, data-driven heuristics, and LLM calls only for missing fields. Includes adaptive SQLite caching, a Flask API, and batch/streaming endpoints.

A Java benchmarking suite built with JMH to measure the performance of core data structures across access, insertion, removal, search, and sorting operations, comparing empirical results against their asymptotic complexity.

A Jupyter notebook exploring credit card fraud detection, covering exploratory analysis and class-imbalance handling, feature preprocessing, and training and evaluating classification models with precision, recall, and ROC-AUC.

I was democratically elected to lead a municipal youth citizenship council, where I ran public debates and elections, founded the council's digital communication channels, represented it in events and municipal media, and took part in an intersectoral commission working toward the municipality's UNICEF Seal.

I supported a national rural microcredit program in partnership with Banco do Nordeste, presenting services to small-scale producers, managing client documentation, and helping automate and streamline the team's workflow.

I organized events and onboarding activities to welcome and integrate incoming Computer Science students into the academic community.