Design a streaming embedding-based classifier

You are building a production system that must classify an unbounded stream of raw text messages in real time. The end-to-end pipeline should: ingest and clean the streaming text; split it into meaningful units; transform those units into dense vectors; incrementally learn a classifier from the vectors; and return predictions with minimal latency. Describe and justify your decisions for each stage: data cleaning/preprocessing, tokenization, embedding construction, model…