This project is part of the field of computational immunology, with a focus on improving machine learning (ML) models for predicting MHC-peptide binding — a key step in the immune system’s ability to detect harmful cells. Accurate prediction of this binding is essential for applications such as cancer immunotherapy, vaccine development, and the study of immune-related diseases. The project explores various embedding strategies for HLA and peptide sequences, and evaluates the performance of multiple machine learning models. It also uses conformal prediction to estimate how confident the model is in its outputs. This work is in line with current research trends that apply ML to biological data, especially where prediction accuracy and reliability are critical. It addresses the need for faster, more scalable alternatives to lab-based binding assays, making it valuable to the biopharmaceutical industry. The project is conducted in collaboration with NEOGAP Therapeutics AB, a Swedish clinical-stage biotechnology company specializing in personalized cancer immunotherapies, supervised by Alejandro Fernandez Woodbridge from Neogap, and supervised by Jens Lagergren from KTH.