Atmospheric Plasma Spraying (APS) is a versatile coating technology with diverse functional features. Deposition efficiency (DE) is a major performance measure in APS, influenced by various factors. Due to intricate interdependencies of these factors, enhancing DE has always been a challenging task in the process development of APS. Hence, employing a variety of computer-aided methods is essential to understand and manage these correlations. The concept of the so-called Digital Shadow combines domain-specific models with data-driven techniques of Artificial Intelligence (AI), inferred by autonomous agents to create a sufficiently accurate image of the production process including all relevant data. This dissertation is devoted to the development of the primary steps towards a Digital Shadow in APS with the ultimate goal of improving the process efficiency. Modern AI methods, namely Support Vector Machine (SVM) and Adaptive Neuro-Fuzzy Inference System (ANFIS), were used in this work to predict DE. To tackle the problem of insufficient data for training the aforementioned AI models two approaches were pursued: 1) A method was developed for in situ determination of spatially resolved deposition efficiencies on the substrate, namely Local Deposition Efficiency (LDE). By using LDE, sufficient amount of data for learning algorithms could be generated, while providing that much data for ex situ measurements of global DE and their corresponding particle properties would be impractical. 2) Simulation data for the in-flight particle properties were generated by using the simulation models of the plasma jet already developed at IOT. The combination of these two strategies provided the aggregated and purpose driven data sets required for a Digital Shadow in APS.