Prof. Dr. Viktor Pocajt and Daniel Trost from Total Materia AG in Zurich, Switzerland, presented on a universal machine learning system for predicting material properties. They discussed the critical need for accurate material property information in engineering simulations and the challenges in sourcing these data from standards, books, articles, and the internet. The presentation highlighted the problems faced due to the lack of knowledge about material properties and standards, leading to engineering errors, missed opportunities in design optimization, and difficulties in international communication between procurement and production. The typical data available for materials like aluminum, stainless steel, and carbon steel were discussed, emphasizing the variability in properties depending on factors such as heat treatment. To address these challenges, Total Materia AG has developed a curated database application like TM Horizon, featuring functionalities like smart comparison, export capabilities for various CAE formats, cross-references and equivalents, advanced search and comparison, compliance assessment, and a material console. However, they acknowledged that it's technically impossible to have all properties for all materials, heat treatments, product forms, working temperatures, and environmental conditions from experiments and standards. The solution proposed is a universal machine learning system for predicting material properties. This system works on a large dataset (over 500K materials and 30M+ data points) using the company's taxonomy and VDA-231 classifications to create separate models for different types of materials, conditions, and properties, resulting in over 1000 ML models. The system uses inputs like material designation, chemical composition, or family, and optional factors like product form and heat treatment. The outputs include mechanical properties at room temperature, temperature dependence of mechanical and physical properties, nonlinear properties, and environmental impacts. The machine learning models are benchmarked with large test sets for the highest result certainty, typically over 90%. Several use cases were presented, including filling data gaps for materials like aluminum and titanium, conceptual design applications, and exploring material variations. The impact of composition and heat treatment on properties like tensile strength and yield strength was also showcased. The integration of this system into the CAE process was emphasized, stressing the importance of saving, reusing, and sharing data to reduce the time spent searching for information, repeating tests, and underutilizing test data.