This document discusses using recurrent neural networks (RNNs) to estimate air quality and atmospheric chemical composition. RNNs can provide estimates much more cheaply than traditional chemical transport models run on supercomputers. The presented Chemistry Estimating Recurrent Neural Network (CERNN) model takes in meteorological data and outputs estimates of atmospheric compounds like ozone and ethane that match estimates from chemical transport models. While CERNN requires large amounts of training data, it can still provide approximations of air quality changes during the COVID-19 pandemic lockdown period when direct chemical modeling data is limited.