The generation of appropriate parameter configurations for high-dimensional hierarchical spiking neural network models is currently one of the slowest and most frustrating tasks in the computational neuroscience workflow. Moreover, many of the currently used methods for multidimensional parametrization only provide single-point estimates, which do not allow deeper insights into the parameter regions of interest and parameter interactions. Having better knowledge about this would allow neuroscientists to understand more about the network dynamics and the relationships between structure and function in the brain. To overcome this limitation, we apply inverse approaches to infer parameter regions which enable the network to reproduce statistical features extracted from experimental data or user-defined features of desire. With the impressive growth of deep learning capabilities in recent years, simulation-based inference (SBI) methods have evolved rapidly [1]. These methods do not need to evaluate a likelihood function, instead they only have access to the forward simulator. The idea in SBI is to first generate a dataset by drawing parameter sets from a user-defined prior distribution and using them to simulate network dynamics. Then, a neural density estimator (such as a normalizing flow) is trained on this dataset to approximate either the posterior, likelihood or likelihood-to-evidence ratio. Having this, the entire posterior distribution can be explored. We integrate the SBI approach into our L2L framework [2], a framework for gradient-free optimization and meta learning. The implementation follows a two loop approach having the inference method in an outer loop and the spiking neural network model simulations in an inner loop. The communication between outer and inner loop is reduced to a minimum, so that the inference method is agnostic to the underlying neural network model. By evaluating multiple parameter sets in parallel, this optimization method can be employed on high-performance computing systems in an embarrassingly parallel fashion. SBI has previously been applied to neuroscience models and neuronal networks. Our particular focus is on inferring parameters regarding the connectivity between clusters of neurons in the network, such as connection strength, synaptic delay and existence/probability of connection. In addition to this, by integrating SBI into our L2L framework, we make it easily accessible, and easy/fast to use on HPC systems.