This talk will focus on learnings gathered when building an enterprise search platform with multi modal content - ranging from highly domain specific content to images to unstructured content. Problems of extraction, inference and relevance shall be discussed, while showcasing cross domain search at scale.