How do you handle missing or highly unstructured data when building a pipeline for analysis at Verily?
