Understanding the ocean is critical for several domains, including the sustainable management, development and promotion of the blue economy (e.g., the exploration of natural and mineral resources), maritime spatial planning, shipping, national security and predict weather and climate. The propagation of acoustic waves within the ocean to study its internal structure emerged in the late 1970s as a promising method for long-term ocean monitoring, leading to the development of a new discipline called ocean tomography. Conventional ocean tomography methods are based on active broadband sources that generate low frequency signal (<50 Hz) that is received meters and kilometres away by acoustic receivers (i.e.,hydrophones). Most application examples deal with the spatial prediction of ocean temperature (or sound speed) at, or above, the mesoscale (~100km) allowing the characterization of horizontal layers of the ocean. Imaging and modelling the ocean sub-mesoscale structure and below (<1-10 km) and the identification of moving, anthropogenic targets, which is critical for the national security, requires the development of new solutions able to perform ocean tomography with higher frequencies and consequently higher spatial resolution. The TRAINEE project proposal aims at pushing the current boundaries of this exciting research field forward and pave the way to develop a new system to perform real-time highresolution ocean tomography using autonomous underwater vehicles (AUVs) and moored hydrophones. The project proposal is divided in two main steps. The first is to create a high-resolution digital twin of the ocean with detailed spatiotemporal distribution of ocean temperature, and sound speed, mimicking real ocean conditions where a field experiment will be deployed. This computational playground will also comprise a numerical wavefield propagation of ultrasonic waves in maritime environment to emulate the positioning of ultrasonic emitters and receivers in multiple locations around targets (Task T2). The TRAINEE digital twin will serve as bed test for the development of novel deep learning methods to process the data online and for ultrasonic tomography based on deep physics-informed neural networks (tomoPINN). After training, deep learning methods will allow its execution with low computational costs, in real-time within autonomous systems. After assessing the advantages and limitation of tomoPINN and the requirements in terms of number of ultrasonic emitters and receivers as a function of the target to be imaged, the second part of TRAINEE consists of a field experiment with acquisition of real data and their analysis and interpretation (Task T3-T4). TRAINEE’s expected outcomes comprise a detailed description of the requirements to perform ultrasonic ocean tomography for maritime surveillance and a prototype system ran in a field experiment. To reach these objectives, the TRAINEE consortium joins a multidisciplinary and complementary research team formed by geophysicists, oceanographers and engineers.