Watch this video on YouTube
You have trained a model using data that was preprocessed in a batch Dataflow pipeline, and now you need real-time inference while ensuring consistent data preprocessing between training and serving. What should you do?