Non-uniform in-mould solidification in long and flat products can lead to defects (e.g. rhomboidity, bulging, …) and/ or surface defects (e.g., cracks) affecting the as-cast quality and the continuous production flow up to rolling, especially when aiming at improved energy efficiency by hot direct charging of billets or slabs to reheating furnaces. This proposal is aimed at an early detection of defects and to identify, with the support of in-line sensors and models, incorrect process conditions which lead to shape or surface quality problems in the as-cast products, that may also lead to problems during rolling. The occurrence of defects will be predicted by means of machine learning and other AI tools. Appropriate countermeasures and suitable operating conditions for defect avoidance will be defined and provided via alerts and recommendations will be developed and applied in industrial campaigns to evaluate the performance of the developed solutions for defect avoidance.