Hunch is a NY startup building a taste graph that will allow them to offer precise, intimate recommendations to users almost instantly.
When users first go to Hunch they answer a series of questions. At the end, the site delivers recomendations to them about everything from poetry to food to clothing.
To do this, Hunch has to compare a user’s answers to more than 30 billion other interactions between people, places and things they have recorded in their system.
That’s a pretty intense calculation to pull off, and Hunch has purchased and perfected a very special machine to do it.
As Peter Warden at ReadWriteWeb details, Hunch decided against the current trend, which is to buy an army of cheap servers and network them together to solve big problems.
Instead they went with a single super powered server which the Hunch team tweaked to optimize it for TasteGraph calculations.
“I have to admit I was surprised that he was apparently swimming against the tide by working within a single uber-machine rather than using an army of dumb boxes,” wrote Warden, “but as [Hunch] explained their requirements it all started to make sense.”
The Observer won’t pretend to understand the finer points of IT and server deployment. But its cool to see a NY startup going against the grain and building a unique machine to power their platform.
bpopper [at] observer.com | @benpopper
Follow Ben Popper via RSS.