Machine learning has been long adapted in many areas of industry such as finance, e-commerce, and marketing. Recently, it is becoming increasingly popular in the physical sciences as well. In particle physics, CNNs are used to help identify rare neutrino events in detectors. In fluid mechanics, neural networks can be used to reduce error and accelerate convergence time when simulating flow fields. However, when it comes to these physical processes, introducing traditional machine learning algorithms to analyze the data will not suffice. While state-of-the-art machine learning models are sometimes able to outperform physics-based models when given a large amount of training data, they can produce results that are physically inconsistent due to their simplified representations of the processes. This talk will dive into current research aimed to improve the modelling of physical processes using machine learning. Different “physics-guided” models will be explored for different neural networks, namely recurrent NNs, convolutional NNs, and feed-forward NNs.