Create nightly sleep records for analytics team before training

Your sleep device records arrive with inconsistent timestamps and missing fields, causing gaps in daily model inputs and delayed insight delivery. Data teams get normalized, timestamped rows for model training before the next run.