Guys, I am real sorry about such an enormous delay in my reply, but evidently I was having some issues with my emails. Hopefully, it is all under control now and I am back in the saddle.
So, I do like the idea: this whole modelops thing goes hand in hand with the original intent of the project. And in fact would only enhance its capabilities. If the community is behind this - not only mentors, of course - it could be done relatively easily. I think it is a matter of adding the code into the git and, perhaps, issuing some additional SGA (in case the code belongs to a commercial entity right now). Could you please tell us a little bit more about the situation: is there a code exists for this new functionality? Do you want to bring something in or the new facilities would be developed from scratch? I am sure we all need some more details to move forward. Thanks! Cos On Mon, Jun 03, 2019 at 11:19AM, Dmytro Liaskovskyi wrote: > Hello everyone, > > As you know DLab is currently incubating. > The podling positions DLab as an exploratory, r&d environment where data > scientists can create and train models. > In order for the model to be productionalized, additional steps needs need > to be made to build and deploy such ML models to production environments at > runtime. This major functionality is not yet part of DLab. > > During DLab incubation at Apache, EPAM developers have been implementing a > tool, which significantly extends DLab with automated model deployment > capabilities, including, but not limited to model performance monitoring, > feedback loop, lifecycle management. > This will bring DLab to a new level – a powerful self-service AI platform > for end-to-end lifecycle management of ML models and environments for them. > > We’d be happy to extend DLab with such capabilities and share it with > community. > > Dear mentors, could you please assist us, defining next steps for podling to > proceed and being able to smoothly integrate/extend DLab with the tool > described above within current DLab incubation process. > > Thanks in advance for your assistance. >
signature.asc
Description: Digital signature