Alzheimer's Disease (AD), a complex neurodegenerative condition, demands precise monitoring and analysis across its severity spectrum, particularly at early stages like Mild Cognitive Impairment (MCI). In response to this need, our study leverages the innovative machine learning approach OPTIVIS to construct Digital Twins for AD patients. These Digital Twins, built on OPTIVIS algorithms, are virtual replicas of patients, encapsulating detailed clinical outcomes based on initial patient data and standard care scenarios. Our approach utilizes a comprehensive dataset from both observational studies and control arms of clinical trials, marked by a diverse array of both present and missing data points—a common challenge in AD research. Through a newly developed OPTIVIS model architecture, we adeptly handle this data variability, offering a robust tool for predicting disease progression. The efficacy of OPTIVIS is validated against an external test set, illustrating its remarkable capability to accurately reflect the progression of critical endpoints in clinical trials for a range of AD severities, from MCI to mild-to-moderate AD. This work not only enhances our understanding of AD progression but also opens new avenues for optimizing clinical trial strategies and patient care.