Gabriel Aguiar

Background

About Gabriel Aguiar

I am a Data Scientist with experience at the intersection of industrial process optimization, statistical modeling, and machine learning. My work spans from developing predictive models and building robust data pipelines (Kedro, PySpark, scikit-learn) to applying interpretability methods such as SHAP to understand model behavior in real-world production settings. I’ve contributed to high-impact projects in mining and banking, including pelletizing optimization at Vale, reagent dosing optimization at Mineração Vale Verde, and current initiatives at Banco do Brasil. My focus has consistently been on bridging advanced modeling with operational decision-making, where constraints and trade-offs are as important as prediction accuracy. Alongside applied industry work, I am deeply interested in the research frontier of model interpretability, particularly in large language models. I see this as a path not only for transparency and accountability in AI but also for unlocking new ways of aligning models with human goals. This is the foundation of my purpose in pursuing a master’s degree, where I aim to explore innovative approaches that combine interpretability, optimization, and human-centered AI. Curious, rigorous, and motivated by challenges, I thrive in environments where data-driven insights meet complex decision-making. 10× faster pipelines: 20-min notebook → 2-min Kedro com ~40 restrições (Vale).+20% customer retention via targeted analytics (Injet Care). 500+ Q&A and 10+ Kaggle competitions delivered (TripleTen).