Agent-Based Models (ABMs) are increasingly being used to model interactions and dynamics within social systems. Consequently, they are also considered to be a very suitable tool for studying the dynamics of school choice and school segregation. However, empirically calibrated ABMs have received limited attention in this field. Mostly due to computational challenges (it is often computationally expensive to run an ABM), a lack of data, and shortcomings of the methodologies to calibrate them. Additionally, the privacy of possibly sensitive data needs to be taken into account. Hence, it remains an open question what direct calibration of an ABM on household-level data can mean for our understanding of school choice and it would make ABM more suitable to represent real systems and to perform policy analysis. Fortunately, ODISSEI provided us with the infrastructure which allowed us to run our computational models at scale while preserving privacy. This has resulted in the ability to run our ABM millions of times on a supercomputer, employ a computationally expensive calibration algorithm and using household-level register data from Statistics Netherlands. Moreover, this study therefore not only provides a methodology to calibrate ABM for the specific field of school choice, but also for ABMs in other application areas, which is an important step towards more empirically realistic ABMs in general.