Materials for the next-generation batteries must ensure high performance at the device level while featuring reduced environmental impact and enhanced safety over the materials’ lifecycle. It is possible to define narrow performance-related tasks associated with individual battery components and address them separately using, for example, physics-based modeling or data-centric statistical learning. However, the patterns of discovery associated with individual device components are different from the patterns targeting the full range of tasks, all the way to the device performance. This talk discusses the discovery pattern in the case when the battery technology and underlying chemistry is significantly novel, and improvement of the device performance is the goal. Artificial intelligence (AI), machine learning (ML), and natural language processing (NLP) play prominent role as components of the discovery loop. In the adopted approach, they complement subject matter expert (SME) knowledge captured via Expert-in-the-Loop (EITL) approaches. We will focus on the case of finding novel electrolytes for heavy-metal-free batteries and describe application of the intelligence augmentation (semantic embedding paired with EITL) and sequential prioritization of experiments to improve cycling performance.