The engineering problem here is entity resolution at city scale. The city publishes a tax lot file, a building footprint file, an address file and a register of deeds, and none of them shares a key with the others in a way that always works. A footprint has to be matched to the lot it stands on, an address point to the structure it names, and a deed to the place it conveyed - and every one of those joins has a long tail where the honest answer is to refuse the match and log it rather than to guess.
The front end is the other half. A directory of 995,043 pages that each read as a finished page, a map that stays smooth on a phone while serving precomputed vector tiles, and a URL scheme with no opaque id in it, which means a wrong borough has to become a redirect rather than a dead end. None of that is exotic; all of it has to be right on every one of the pages.
It suits people who would rather be correct slowly than confident quickly, and who find it satisfying that the coverage page states where each dataset is wrong. If that sounds like the job you want, say so even when nothing is posted.
