Criteo Data Preprocessing & Log-Binning Strategy

Describe your strategy for efficiently handling the large Criteo dataset. How will you load and preprocess the raw data (categorical and numerical features)? Detail the log-binning strategy for numerical features, explaining how you will convert dense numerical values into categorical embeddings. Also, specify how sparse categorical features will be handled (e.g., hashing trick, embedding layers, handling OOV values). Provide a high-level data pipeline design, including batching and efficient loading.