Edwin Zajac

Built relational database with automated SharePoint API synchronization for real-time data ingestion from CFD results. Developed energy consumption prediction models for electric boat fleet using ensemble architectures, achieving 8% MAPE for ship resistance calculations and reduced worst-case scenarios. Deployed Streamlit application enabling non-technical staff to generate energy forecasts, reducing CFD simulation time for pre-project calculations. Implemented model validation framework and conversational AI interface for dataset exploration.

Worked on QPSK modulation and computed encoding/decoding error in message transmission based on a paper research, including a study of numerical stability and estimation of key parameters.

Developed a semi-supervised learning method for the recognition of diseased apple tree leaves, combining YOLOv7 for detection and classification algorithms like Efficient Teacher on 3000 images, with a comparison of two approaches to optimize performance.