Quasar-o-meter: A Classification Model for Radio Signal Processing

Last spring, I did an internship where my team worked directly with the US Space Force. Our task was to come up with a way to easily distinguish solar radio bursts from other radio waves, given how quiet the Sun is compared to other radio sources like a cell tower. Such a project instantly spoke machine learning and data science to us. So we trained a convolutional neural network (CNN) on synthesized spectrograms as our data, and the model had over 90% recall when predicting the different types of solar bursts, if there were any, as well as when there's radio frequency interference (RFI).