Projects

This research-focused project explores advanced time-series forecasting techniques for modeling the highly noisy and irregular emission patterns of Active Galactic Nuclei (AGN). The study compares classical machine learning baselines, LSTM networks, and transformer-based large time series models (Sundial) using multi-resolution data (daily, weekly, and monthly). A censoring-aware preprocessing pipeline was designed to handle upper-limit observations, noise, and data sparsity without target leakage. The model incorporates object and granularity embeddings to enable global learning across multiple AGNs while preserving source-specific behavior. Experimental results show that LSTM models outperform transformer-based approaches on limited datasets, highlighting important trade-offs between model complexity, inductive bias, and data scale in real-world scientific forecasting tasks.

This project focuses on multilingual, hierarchical text classification for narrative analysis, designed to categorize news articles into high-level narratives and sub-narratives across multiple languages. Using RoBERTa for English and XML-RoBERTa for Portuguese, the system applies multi-label, multi-class classification to detect and analyze propaganda, misinformation, and media framing. By implementing data balancing techniques such as oversampling rare classes and threshold optimization, the model significantly improves multi-label detection accuracy. The pipeline integrates rule-based filtering, transformer models, and fine-tuned classification strategies, ensuring robust and scalable AI-driven analysis. This work achieved 10th place worldwide in a competitive Samevel Challenge , with F1 scores of 0.70 for Climate Change sub-narratives, 0.69 for high-level classification, and 0.64 for Ukraine-Russia War sub-narratives.

Durdo is a cross-platform mobile application developed using Flutter, designed to empower users in tracking their activities and engaging in various sports competitions. As the lead full-stack developer, I spearheaded the end-to-end development process, leveraging Flutter for the front-end and integrating robust cloud-based solutions like Google Cloud and Firebase for the back-end infrastructure.The application's user interface was meticulously crafted in Figma, ensuring an intuitive and visually appealing design that enhances user experience.