In fusion reactors, large numbers of high-energy neutrons are generated, creating a harsh and demanding environment for reactor materials and components. In particular the ITER radiation environment will be characterised by harsh conditions in terms of neutron and gamma fields: consequently it is crucial to test its sensitive components, such as electronics, in dedicated facilities. To address this need, we propose the development of the GENeuSIS (General Experimental Neutron Systems Irradiation Station) project. GENeuSIS is an innovative test facility designed to assess and characterise the behaviour of diagnostics, electronics, and other ITER critical components when exposed to 14 MeV neutron irradiation from the Frascati neutron generator. The GENeuSIS assembly consists of a set of neutron-moderating materials slabs enclosing an inner cavity where the neutron and gamma spectra foreseen in specific ITER locations are reproduced. In this context, a machine learning (ML) model automatises the selection of materials required to achieve the desired neutron and photon spectra. This work focuses on the development of a supervised ML model, specifically a neural network, trained on a database generated from previous neutron transport simulations using the MCNP code. These simulations have already demonstrated the feasibility of GENeuSIS by replicating the neutron spectrum at specific ITER locations, such as the Port Interspace (GENeuSIS-I assembly) and Port Cell (GENeuSIS-II assembly). However, the design of each GENeuSIS assembly using Monte Carlo methods generally demands significant computational resources and depends on extensive ‘trial-and-error’ transport simulations, often resulting in a slow process. The proposed ML model aims to accelerate the optimisation phase of GENeuSIS assemblies by rapidly identifying promising configurations, which are subsequently validated through full Monte Carlo simulations.