In this work, we study the automated detection of exoplanets, planets that orbit stars other than our sun. A range of techniques exists for exoplanet detection, from astrometry to radial velocity, to the common transit method, which plots the relative brightness of an exoplanet’s host star to make a light curve. However, given such a light curve, determining if an exoplanet exists is nontrivial, as a dip in brightness can come from various sources, such as noisy observations or extraneous systematics like cosmic rays, atmospheric interference, noisy light sources, etc. Recent work has shown that deep learning techniques from large labeled datasets are promising ways to automatically predict the existence of an exoplanet, where a neural network is trained to take in a light curve and output a probability that it came from an exoplanet. In this study, we compare a number of machine learning architectures, including convolutional networks and sequence-based models to identify which model is most effective at identifying exoplanets from phase folded light curves. We test these architectures on a synthetically generated dataset mimicking TESS observations from a single sector, and evaluate which models have the highest accuracy on unseen light curves.