In the past decade, the concept of brain age has been introduced as a biomarker for quantifying age-related morphological change in the brain. Commonly, machine learning is used to predict age from neuroimages. Brain age is potentially a useful biomarker in 1) clinical trials evaluating putative neuroprotective interventions and 2) clinical practice to assess the risk of age-related brain disorders on an individual level. However, while the models developed so far have been accurate in predicting chronological age, their validity has been questioned, as they show weak predictions of future age-related structural changes in the brain. Here, I propose to utilize recent advances in machine learning, specifically self-supervised learning, and develop a model of aging trained on longitudinal data, aiming to estimate healthy as well as accelerated aging rates reliably. Having a reliable model of aging trained on the entire lifespan will aid in in detecting early deviations indicative of, e.g., future onset of neurodegenerative disease. By detecting such risks on an individual level, early interventions, as well as more precise monitoring, can be put into place, with the potential of preventing the onset of age-related morbidity. The first stage (2024) will be used to implement different architectures and familiarize myself with the workflows. If appropriate, we might apply for MEDIUM/LARGE allocations in 2025.