You are tasked with developing a custom classification model using TensorFlow, based on a large tabular dataset stored in BigQuery. This dataset contains hundreds of millions of rows and includes both categorical features, such as SKU names, and numerical features. As part of your preprocessing steps, you must apply a MaxMin scaler to some of the numerical features and use one-hot encoding for some of the categorical features. Your goal is to train this model over multiple epochs efficiently, while also minimizing the effort and cost associated with your solution. Given these requirements, what approach should you take? | Google Professional Machine Learning Engineer Quiz