In this project we are developing an AI accelerated self-driving lab (SDL) for novel materials exploration using physical vapor deposition (PVD) methods. The SDL integrates hardware automation, automatic data processing and AI driven decision making to efficiently navigate the vast process parameters space and efficiently choose the next point of experiment. Given the time consuming nature and complexity of PVD experiments, our goal is to maximize learning while minimizing the number of experiments conducted. For this task we use Bayesian optimisation with Gaussian process (GP) regression. We investigate different GP models and acquisition functions to fully characterise PVD processes to obtain a certain target composition. This can lead to finding new materials much faster compared to a traditional approach.