I don't work in Uber's team on ML, so I can't answer this, but I also cannot just simply assume everything is built-for-PR-splash. You can argue why can't they simply talk about the reason of not reusing any existing OSS framework than building yourself, but making an assumption that they did it for the sake of PR is lack of factual support. Are we now living in a world assuming everything Uber did was for PR? Last I heard they had 1500+ engineers and with that size, I am sure they can just talk about things they have done in the last few years.