The recent economic crisis has also been a crisis for economic theory. Hence, the economic profession started to debate and prepare a methodological reconsideration about the general role of economic models; new modeling strategies, alternative and complementary to the standard Dynamic Stochastic General Equilibrium (DSGE) models, have therefore emerged. Among these innovative approaches, Agent-based Computational Economics (ACE) started to occupy a prominent role. However, ACE models still need to be improved in three main directions before being ready to be employed as a standard policy tool adopted by central banks. These three directions are: (i) calibration, (ii) external validation, (iii) formal evaluation of systemic risk. This fellowship aims at directly tackling these three issues by employing new statistical learning, econometric, and algorithmic techniques and by applying them to an ACE model that enables one to analyze private and public debt dynamics by closely following the financial and real sectors at the micro-level. The research plan here proposed aims at developing a calibrated and validated model able to explain how the rise in private debt and the interconnectedness of financial institutions might lead to financial crisis, which then might spread to the real sector and ask for a massive public sector intervention, possibly generating also public debt overhangs. Additionally, the innovativeness in the methods here employed will allow to establish higher standards for the development of descriptive ACE models. Once these model are properly and rigosously calibrated and validated by means of real-world datasets indeed, they can be adopted to evaluate a set of counterfactual policy exercises. In particular we aim at understanding which ex-ante policy measures might help avoiding debt-triggered crises and which ex-post policy interventions might help in mitigating their negative effects.