Phosphor-converted white light-emitting diodes (pc-WLEDs) offer energy-efficient and eco-friendly means to address lighting requirements in a wide range of settings. Machine learning (ML) has proven to be good at handling large quantities of data. Over the last few decades, substantial data has been generated on phosphors for pc-WLEDs. However, this data has neither been collated nor used in ML platforms towards the development of novel phosphors. In this context, we indicate that ML techniques can be used to identify the compositional spaces ideal for the phosphor design (this would be the use of an ‘inverse problem’ approach in phosphor design). As a case in point, identification of search spaces using CCT 70 as features is shown in the article. This generates further prospects for future work.