In recent years, there has been particular interest in the raw materials required for lithium-ion batteries: nickel, cobalt, manganese, graphite, and lithium. Demand for these raw materials is expected to grow markedly in the coming decades, providing a potential economic opportunity for those developing countries with appropriate resources. While there is an obvious need for geologists, mineralogists, metallurgists, economists, and environmental experts to work together to identify and develop the best resources of all these raw materials, artificial Intelligence (al) is playing an increasingly important role in rapidly boosting the capability of the geologist, ultimately aiming to efficiently develop orebody knowledge. Machine Learning (ML) is a branch of al that aims to build models to iteratively learn from data, finding insights and structure without being explicitly programmed on where and how to look. This paper describes two types of ML models - segmentation and classification - built to quantify spodumene crystals and classify pegmatite texture using good quality, high-resolution core photography, from diamond drilling on the Goulamina pegmatite lithium deposit. The segmentation model identifies individual spodumene crystals, visible in diamond core photos, providing quantitative information on the size, shape, area and colour - all key to understanding the mineralogy. The classification model defines the pegmatite as coarse-, medium- and fine-grained pegmatite. Both the combined quantification of spodumene crystals, and pegmatite texture classification, provided foundational data sets not previously manually logged by geologists on-site. Indeed, once these models were trained on known examples, they were rapidly applied across diamond core photography taken across the entire deposit. We discuss why and how these models were trained, how the results have improved knowledge of the orebody, and their potential use in resource estimation.