Chromatin is a compressed version of DNA that can be opened or closed, which is important because a gene cannot be expressed when chromatin is in its closed form. ATAC-seq is a sequencing technique used to determine what parts of chromatin are opened or closed. However, the quality of ATAC-seq data is heavily dependent on many environmental factors, and therefore prone to noisiness. Our goal was therefore to create a model to denoise low quality ATAC-seq data. We used a NASH architecture search to develop a CNN architecture that was trained on high quality, low quality pairs of data. Our final model is able to reliably convert noisy data into high quality data.