The quality of the data plays a significant role in the success of a ML model. A well-curated dataset with a diverse range of examples will lead to better performance compared to a dataset with a limited number of examples. It is important to gather a large and representative dataset to ensure the model can generalize well to new unseen data. Unfortunately real data collection process can be expensive and hard to organise. The goal of this talk is to show potential of synthetic data for training deep learning models in two application areas, namely heathcare and mobility.