Michelangelo: Uber’s Machine Learning Platform

Re:Michelangelo: Uber’s Machine Learning Platform by aorloff on Wednesday September 06, 2017 @01:04AM (#15181358)

I am surprised by this - h2o.ai seems to have many if not all of the features of this, and is open source.

"Specifically, there were no systems in place to build reliable, uniform, and reproducible pipelines for creating and managing training and prediction data at scale. Prior to Michelangelo, it was not possible to train models larger than what would fit on data scientists’ desktop machines, and there was neither a standard place to store the results of training experiments nor an easy way to compare one experiment to another."

Seems like h2o.ai fits a lot of that bill.

[ Parent | on HN ]

Re:Michelangelo: Uber’s Machine Learning Platform by opensandwich on Wednesday September 06, 2017 @01:38AM (#15181460)

h2o.ai doesn't really do data pipelines, though it does appear they are eager to go into this space through their new driverless.ai tool. However this does not appear to be open source.

[ Parent | on HN ]

Re:Michelangelo: Uber’s Machine Learning Platform by JPKab on Wednesday September 06, 2017 @12:35PM (#15184971)

Something else that throws me off on this:

My team's experience with MLlib has been bad. Especially compared to H2O Sparkling Water.

Anyone else on here find MLlib to not be as good as advertised? I was surprised to see Uber using it.

[ Parent | on HN ]

Re:Michelangelo: Uber’s Machine Learning Platform by thecity2 on Wednesday September 06, 2017 @02:39PM (#15186038)

Spark is good if it fits your use case (or if you can make it fit your use case). We have definitely found that scaling can be a real issue though.

[ Parent | on HN ]