This MAY look like a normal Jupyterlab, but it's not.
It's running inside one of 4 agents installed on 4 different servers in a google datacenter with access to TPU's in a different country, and neither has public IP. OpenFlow can now expose load balanced ports across agents, and expose them to localhost (given you have the right permission ofc )

One small step, in making scaling and deploying machine learning much easier.
#openflow #openiap #mlstudio #jupyter

I been in crisis the last week. Going down the tensorflow-js route has turned out to be wrong.
I did not like having to give up on all the last 2 months work, but there is no way around it
I need to stay true to the original goal of creating something that makes ML more scalable and easy to integrate with openflow when working with big data sets.
So I'm working on creating a custom jupyter backend that uses OpenFlow for authentication, data and network
#openflow #openiap #mlstudio