Virtual brain twins are personalized digital models of individual human subject or patient’s brains, allowing for mechanistic interpretation of neuroimaging data features. Training & inference with these models however presents a pair of challenges: (a) large shared infrastructure do not allow for use of personal data and (b) inference in clinical applications should not require significant resources. We introduce “anonymized personalization” to address both by expanding model priors to include personalization which under amortized inference allows training to be performed anonymously, while inference is both personalized and lightweight. We illustrate the basic approach, demonstrate reliability in an example, and discuss the impact on both experimental and computational neuroscience. Code is available at https://github.com/ins-amu/apvbt