How would you structure a data pipeline to ensure low-latency inference for financial news at Bloomberg?
