The danger of forest fires has significantly risen over the past decade due to climate change and improper forest management. Wildfires have a severe effect on social and ecological systems. Being able to predict the severity of forest fires would be valuable knowledge to have fighting fires or when allocating resources for forest management. In this work, I attempt to apply machine learning techniques to accomplish this task. Although machine learning algorithms have been shown to be powerful in many other fields, these methods have been underutilized in the prediction of wildfire severity. First, I built the wildfire dataset which consists of various domain features, ranging from land management and logging practices, including logging data, forest composition data, stream locations, stand age index, and satellite imaging. Second, I employed multiple machine learning techniques to predict the severity a wildfire would have on an unburned forest using publicly available datasets and remote sensing data. My models predict the severity class pixel by pixel of the satellite image which can be used to provide fire severity prediction maps. The most promising machine learning techniques were a random forest model, a deep neural network and a 1 dimensional convolutional neural network with the respective accuracies of 66.7%, 61.40 %, and 69.67 %. These results could offer a way to better control and reduce forest fires by helping firefighters fight fires and by predicting future fire severity so that we can target locations for better land management practices.