-Developed causal impact regression models using time series analysis for a
FastAPI web platform (collaborative project), achieving stable confidence
intervals and up to 25% relative impact across multiple datasets.

-Built time series forecasting models (30-day projections) with controlled
confidence bands below 50% of maximum values.

-Developed classification and regression algorithms using softmax classifiers
and single-hidden layer neural networks (with varying neurons and activation
functions), achieving 96-97% accuracy in most exercises.

-Implemented medical image classification algorithms for disease detection,
contour analysis, and color pattern recognition, both with and without machine learning libraries. (Still in progress, degree project)

-Conducted comparative analysis of fully connected neural networks using
linear vs polynomial proposals (up to degree 3), demonstrating improved
accuracy with higher polynomial degrees after network normalization.