Developed the Orbit Angle Prediction Module in the Julek-AK/AE2224-I-B04 repository, establishing a data-driven baseline for orbit angle forecasting using Python and PyTorch. The work included designing a neural network with multiple hidden layers and ReLU activation, implementing a training loop with SGD optimization and MSE loss, and building reusable data-loading and training scaffolding to support rapid experimentation. Additionally, addressed technical debt by removing legacy LSTM prediction logic and updating data interfaces to prepare for future predictor migration. Demonstrated skills in data preprocessing, deep learning, and machine learning, with a focus on maintainability and production-readiness throughout the project.