Palantir is just another Theranos in disguise, it essentially sells an under-performing graph database with outdated visualization tools while making money providing consulting services using "Forward Deployed (OMG!! You mean like soldiers! /s) Software Engineers".
It's a hard interview for us because it's built like a college exam, and how many years has it been since we've taken a college exam? But their target audience, well they just finished midterms.
I was working for a competitor when they tried to get me on board as one of their "forward deployed software engineers" - which appeared to be a lot of big talk for hiring a bunch of recent grads and burning them out pumping out menial interfaces for customers so they would feel they were getting value.
The interview was very normal, not particularly onerous.
The one thing I've heard consistently from the start is that they pay poorly. Generally, the best people don't take low offers.
They could make their interview questions impossibly hard if they wanted, perhaps to add to their mystique. But at the end of the day, they get whoever is willing to take their offer.
Even if you don't have the roughest and toughest devs in the market, surely spending more time with NYPD than the Palantir FDEs (are engineers part of the skeleton staff mentioned in the article?) would get a more valuable result?
Of particular note was when i2 Group took Palantir to court for a conspiracy to misappropriate its software and trade secrets. They settled out of court (for a huge sum of money.) IBM swooped in soon after.
In the past, it lacked the huge amount of transforms in Palantir to get data into the worksheet. Of course these transforms were written by expensive Forward Deployed Engineers to write glue code and convert disparate protocols and file formats. It turns out that doesn't scale well, and it looks like Palantir had to raise prices.
Is anyone surprised that Palantir's spin is about embarrassing levels of image management? That is not just a cultural norm for them, that is their business model.
> The new system, named Cobalt, is a group of IBM products tied together with NYPD-created software.
You really have to wonder about the quality of this internal product. From what I've gleaned about Palantir over the years, they supposedly have one of the hardest dev interview processes. People that work there are on top of their game.
My gut feel is that NYPD does not have as robust of an internal dev situation. Or perhaps they shopped it out to contractors? Either way... skeptical that the system is better. But who knows! It's possible the subset of features a police department actually needs from Palantir might be much smaller than the feature set they're paying for (and not too hard to construct themselves). Whatever is going on, it's clearly a cost-benefit issue.
Although, if they're fighting for Palantir's analysis, it seems they need something that Palantir does that they can't do themselves from the data. There really isn't much one can conclude from this article. Only speculation.
EDIT: From responses on this thread it seems that the "hardcore super secret analysis" image Palantir promotes might be more bark than bite. What if NYPD actually did create something better? Intriguing.