AI & Data Science student at Universidade da Beira Interior, building practical machine learning systems — from computer vision pipelines to NLP classifiers and data engineering workflows.
Alongside my degree, I work professionally as a Video Editor for Wuant, one of Portugal's largest content creators, where I apply the same attention to detail and deadline-driven mindset I bring to every technical project.
Languages: Python · SQL
ML & AI: PyTorch · HuggingFace Transformers · scikit-learn · OpenCV
Data: pandas · Plotly Dash · SQLite
Tools: Git · Google Colab · Gradio
recycling-classifier-resnet50
Computer vision system classifying 7 types of urban waste containers. Tackled severe class imbalance (46% majority vs <1% minority) using ResNet50 transfer learning + Focal Loss. Achieved 87.9% accuracy and F1=0.88. Includes Grad-CAM explainability.
emotion-classifier-bert
Fine-tuned BERT for emotion detection in song lyrics across 6 classes. Trained on GPU (T4) via Google Colab. 94.4% accuracy. Deployed with a Gradio web interface supporting Portuguese/English input.
worldbank-etl-dashboard
End-to-end ETL pipeline extracting World Bank economic indicators via API, transforming with pandas, loading into SQLite, and visualising with an interactive Plotly Dash dashboard.
ml-interpretability-xai
Implementation of IROF and Input Invariance metrics to evaluate and compare XAI methods on ML classifiers, using the Quantus framework.
