Material qualification is an important pre-requisite for design substantiation of any power plant. Historically, this is achieved through large experimental programmes that are eventually collated to support design standards (e.g. ASME) or later in assessment codes (e.g. UK’s R5 and R6). This process is slow and expensive but low risk. In parallel, computer simulations have expanded their roles in the design and assessment process. Advanced physics-based simulations techniques such as crystal plasticity frameworks are increasingly being used to inform the engineering practices. However, they require extensive research to validate and substantial training for the practitioner to ensure the validity of their results. They are therefore considered to be expensive techniques that are deployed at exceptional circumstances. In this paper, a road map to use recent advances in machine learning is proposed that can simplify the complex physics-based simulations and produce high fidelity surrogate models that can be used cheaper, faster, with less stringent training. The surrogate models, because are based on rigorous physics-based simulations, can form part of the material qualification thus accelerating the process and making it more efficient.