Data Scientist & AI Engineer CESAR
Transformed unstructured safety reports for Petrobras into predictive risk intelligence dashboards using NLP, ML classification, and RAG architecture (Python/LLMs), enabling technical teams to proactively identify and prevent workplace hazards through precise data analysis and model outputs, with a strong focus on data quality, curation, and adherence to classification rubrics for language models. Engineered a Computer Vision (OpenCV) lip-mapping system for the SXSW 2025 Innovation Award-winning 'Smart Lipstick' for Boticário Group through extensive dataset construction and enrichment, enabling visually impaired women to apply makeup autonomously 3x faster by ensuring the highest precision in data annotation and segmentation. Designed a Dockerized predictive maintenance system to anticipate machine failures for Ball Corporation, employing comprehensive supervised (Regressions, Decision Trees, Random Forest, XGBoost) and unsupervised learning (Clustering, PCA), achieving a sustained 2% local production increase over five years, with an emphasis on validating and maintaining the quality of input data.