Interface segregation has a pronounced impact on the properties of Ni and its alloys. One of the prominent examples is the hydrogen-induced intergranular material failure also known as the hydrogen embrittlement (HE) that occurs in practically all known metallic alloys upon exposure to H-containing environments. Grain boundary segregation plays one of the key roles in this process and has been considered by many as an efficient way to influence material susceptibility to HE. In this work, we would like to present a predictive multiscale modelling approach for evaluation of the GB chemistry and cohesive strength based on DFT data. We use both phenomenological and machine-learning types of models to assess the interfacial excess of the alloying elements and their impact on the GB properties. The efficiency of the approach is demonstrated by its application to model-based design of a new HE-resistant Ni-base alloy.