We present an ab initio modeling approach for graphene oxide (GO), augmented by machine learning based structural simulations, to extract the optical and electrical properties of the material at different scale levels. The approach addresses a longstanding challenge in GO modeling, crucial for the domain of sensor: capturing the impact on electrical properties provided by variable amount and typology of oxygen functional groups, complex amorphous structure and molecular adsorption of gases and molecules. We show the validation of our model against experimental absorption spectra and photoluminescence data obtained during laser irradiation processes, and we discuss its extension to simulate electrical properties with larger cells of tens of nanometers size, obtained with machine learning interatomic potentials, following preliminary tests we conducted. Breaking with conventional approaches, our framework has the potential to enable electro-optical property predictions on large scales under varying physicochemical scenarios. Those include variable oxidation level, pore formation, and binding with supramolecular species, all of which are common in sensor fabrication processes or during the sensing itself. While this is, indeed, of paramount importance for the mentioned field, further domains of application could be environment, optoelectronics, energy storage, and microelectronics.