The topside ionosphere is the gateway between the lower ionosphere and the plasmasphere/magnetosphere above, and
is critical to the function of GPS/radio communications, satellites, and the power grid. Existing models of the topside
ionosphere are empirical or use empirical model as input, and not as resiliant to extreme conditions, like the effects of
solar storms. In this thesis, I investigate the use of machine learning to develop a model of the topside ionosphere that is
built on in-situ satellite data and is better suited to accurately modeling the ionosphere in extreme conditions, develop
extensions of the model by designing a hybrid model that blends empirical models with machine learning allowing for the
advantages of both types of models to be combined, provide information about relative model importance within a hybrid
model, and use the improved topside models to improve total electron content modeling. I begin by investigating the
intersection between modeling the Earth's ionosphere and using machine learning to model systems in Earth's upper
atmosphere, which illustrates the need for a better topside ionospheric model. This leads into the development of a neural
network model of the topside ionosphere, including the feature selection process and performance analysis of this purely
machine learning based model. To expand the model domain, I apply stacked generalization to combine the developed
machine learning model with existing empirical models, specifically the International Reference Ionosphere and E-CHAIM,
and determine model importance within the stacked generalization model using Shapley values. Finally, I analyze the use
of topside electron density predictions to improve total electron content modeling, and provide better TEC predictions.