Sort of plays right into the whole argument being made over on another thread (https://news.ycombinator.com/item?id=11105198) that companies shouldn't rely too heavily on proprietary tools to solve their data challenges.
Palantir is the poster child for black box software company that keeps trying to sell you more expensive licenses for their mystery algorithm. I know they've had some success in the government sector but I've seen them make a real mess of things in the private sector. At my last company their system just became a giant paracite where data went in and stuff came out the other end but it was entirely unclear what happened in between. The more we looked into the mystery middle ground the uglier it got.
There is no mystery algorithm. It's just cheap college labor working obscene hours to pump out "data analysis" scripts. The sad thing is that it actually works. The places that hire palantir are so mired in bureaucracy and process that whatever palantir builds seems to be better. Of course, just like SAP and Accenture you're now kinda locked in with support contracts.
Modern tools like spark and mesos in theory let you do what palantir does and then some but it assumes your internal IT team is semi-competent which more often than not is not a correct assumption.
Maybe it tops out early, but I'm given to understand junior salaries are also very competitive. (Maybe this is less true by comparison in the States -- it's fairly true here in the UK.)
Palantir is the poster child for black box software company that keeps trying to sell you more expensive licenses for their mystery algorithm. I know they've had some success in the government sector but I've seen them make a real mess of things in the private sector. At my last company their system just became a giant paracite where data went in and stuff came out the other end but it was entirely unclear what happened in between. The more we looked into the mystery middle ground the uglier it got.