Gene therapy is capable of treating diseases that are “undruggable” by small molecule drugs. At the center of gene therapy is the development of efficient and specific gene delivery vectors. For example, adeno-associated virus (AAV) based vectors are able to deliver gene therapeutics to many different types of cells due to their generally broad tropism. Unfortunately, there are some cell and tissue types that are resistant to AAV transduction, and delivery to off-target organs could lead to undesired side effects. Therefore, it would be valuable if we could engineer AAV vectors to transduce specific desired cells and to reduce delivery to off-target tissues. Furthermore, if we could predict how different AAV vectors will perform in different animal models, it would enable us to select the best virus to be used for certain applications. In this work, I have taken a data-driven approach to modify and engineer AAVs to be capable of transducing specific cells. Additionally, I describe a machine learning approach to predicting AAV in vivo biodistribution. Both approaches will help accelerate the design and screening of AAV for future gene therapy applications.