Nicolai Olsen contributed backend engineering to the CogitoNTNU/DeepTactics-Muzero repository, focusing on improving code observability and correctness within the training loop. He enhanced the training loss logging by tracking total policy, reward, and value losses, and introduced optional aggregated loss outputs to support debugging and metric interpretation. Using Python and leveraging his skills in deep learning and debugging, Nicolai fixed a duplicate calculation in the loss function and clarified the initialization and accumulation of total losses with detailed comments. His work strengthened the maintainability and reliability of the training process, ensuring more accurate monitoring and easier troubleshooting for future development.