In terms of a brief overview, this project focused on predicting academic performance from a set of chosen features of multiple modalities. It provided an opportunity to work with satellite images and convolutional neural network architectures to extract useful representations from our data. It further gave an opportunity to think deeply about the role data in any machine learning system and what to do in the face of incomplete data.

You can view an early write-up of this project, which provides additional motivation along with the iterative development of the machine learning model, here.

This section will be updated with a description of the graph neural network approach, but for now, feel free to reach out to me about that.