My Master’s thesis focuses on the development and evaluation of machine-learning approaches for inter-subject fMRI data alignment. Besides conventional anatomical alignment, recent research suggests that mapping neural activity into a shared functional latent space is an effective tool for improving inter-subject data integration. This project specifically explores the utility of variational autoencoders (VAE) as a deep-learning tool for learning shared, generative representations across subjects. Its alignment performance is evaluated against that of an established linear algorithm, the multiset canonical correlation analysis (M-CCA), using a downstream cross-subject decoding task on the learned latent space.