This continuation proposal builds upon the previous project, which focused on simulating and forecasting the flow of goods in a manufacturing supply chain. In this phase, the emphasis shifts towards the application of AI-based irregular time series models. We would like to address the challenges posed by irregular data collection and reporting intervals that are common in supply chain and manufacturing data. We have completed an extensive theoretical review of many models based on Transformers, GNNs or NeuralODEs and we are now heading towards testing these models experimentally. We aim to improve forecasting accuracy on irregular time series. The theoretical groundwork is in place, and we will move forward with experimental application.