For causal estimation in brain pathological conditions, models offer crucial insights into neurophysiological mechanisms. Model-based inference involves constructing a statistical or mechanistic model that captures the essential features of the data-generating process. In this context, simulation-based inference (SBI) using deep neural networks provides an invertible map between parameters and data features. Therefore, there is a need to automatically extract the low-dimensional informative data features, to train the neural networks, the so-called Normalizing-Flows. In this project, our goal is to utilize machine learning algorithms to provide input into the SBI pipeline. The data for this project will be, simulated data from models at different scales and corresponding electrophysiological data such as patch-clamp and intracranial stereoelectroencephalography (SEEG) recordings. The objective is to classify seizure onset and bifurcation patterns from the complex biophysical models. The trainee will benefit from the support of the various skills available within our teams.