The epoch of LSST brings about a step-like change in both data quantity (much greater) and quality (somewhat lower due to the impossibility of spectroscopic follow-up) to SN Ia cosmology, which will correspondingly introduce new challenges in terms of computational and modelling requirements. This chapter reviews the existing methodologies (focusing on templating fitting and summary-based standardisation and cosmological inference) and discusses potential pitfalls in their application to future data (i.e. identifies areas that require improvement, summarised as scalability, statistical rigour, and modelling flexibility). Finally, it presents a grand unified vision for scalable and principled application of neural SBI in the field.