Amorphous zeolitic imidazolate frameworks (aZIFs) are designable nanoporous metamaterials demonstrating a vast capacity for functional stimuli-responsiveness, high mechanical robustness, and large-scale processibility surpassing their crystalline counterparts. Ideally, structure-function relationships would help navigate this huge aZIF design space, just as for crystalline ZIFs. Yet, their lack of long-range structural order makes identifying these aZIFs’ structure highly challenging. In this project, we aim to make a major leap forward by developing two in silico methodologies and combining them in an integrated and generally applicable workflow. First, we will train a ZIF-transcending machine-learning potential (MLP) to model the interatomic interactions accurately. Just like ab initio methods, this MLP will be able to reproduce the experimentally observed amorphisation under heating and pressurisation, but at a substantially lower computational cost. Second, we will develop an in silico nuclear magnetic resonance (NMR) workflow to fingerprint the local amorphous structure of ZIFs and store these fingerprints in an NMR library. This library, in turn, will be adopted to locally map the ZIF structure onto distinct ZIF states. By combining both methodologies, we aim (i) to shed light on the nucleation and growth of amorphous states upon heating and pressurising crystalline ZIFs and (ii) to derive structure-function relationships for aZIFs and enable their functional design.