This document summarizes a data scientist's work using deep learning to classify different types of cries from recordings of her newborn baby. She collected around 70 audio recordings spanning the baby's first 4 months. Using these recordings, she aimed to build models to classify cries as hungry, angry, needing a diaper change, etc. Her best models achieved accuracies between 76-86%. However, she acknowledges limitations from a small dataset and risks of overfitting. She outlines plans to expand the data collection and improve the models to develop a baby cry translation app.