Appreciate you sharing the vision. Having worked in this space for a while, IMO the biggest challenges for a public facing graph are in 1. Entity Linking from NL Query -> Graph queries or in your case relational queries (Multiple similarly named teams in Microsoft). And 2. Relevance of results for more complex queries. I like your approach of having a drop down of filter tags, which eliminates 1, but will be harder to scale like in a Graph of everything.
Great work, and interesting to see Knowledge Graphs in a production setting. Why did you choose a Knowledge Graph as the backend? How is the graph modeled. Do you use existing Graph Query languages, or did you have to create your own?