Leonardo Oliveira

I developed a factor-graph-based 2D LiDAR SLAM system for autonomous mobile robots operating continuously in changing industrial environments. The system combines scan matching, loop closure detection, and pose-graph optimization to maintain accurate and consistent maps during long-term operation despite environmental changes. I worked on improving localization robustness, handling perceptual aliasing and environmental changes, and evaluating system performance using trajectory accuracy and runtime metrics. The solution was implemented in C++ and ROS 2 and designed for reliable operation on production autonomous robots.

I built an Apache Airflow pipeline to collect and process company mission statements at scale for use on a public-facing website. I designed the workflow to orchestrate data ingestion, preprocessing, classification, and storage across large sets of company data. A deep learning classifier was developed to identify sentences representing company mission statements in noisy text sources, enabling automated extraction rather than manual curation. The project combined data engineering, workflow orchestration, NLP, and machine learning in a production-oriented pipeline.