We will discuss the development of new machine learning methods for drug discovery. We will initially focus on an iterative, and active learning powered, approach to optimizing existing large molecule (antibody) leads. We will compare multiple new methods for generating antibody like sequences, and evaluate their utility in a multi-drug-property optimization setting. We will also examine new methods for combining generative AI with protein/molecular structure. Early de novo affinity results will be discussed. Lastly, opportunities for using ML to build frameworks that unite small molecule and large molecule drug discovery insights (multiple modality frameworks) will be discussed.