Hi,
I'm developing different recommendation engines using R. The output from my R code will be passed to a small online shop.
I do not yet have access to the web-app, or the platform where it's hosted. At this moment, I'm just doing some research to better understand my options with respect to implementing my R code into the overall architecture.
At the moment, Docker + Azure Container Services seems like a good match; it would allow me to train and test models locally with my GPU, and deploy / replace models in the cloud. There will be multiple engines, and they should be easy to replace when
a better model architecture is discovered.
Do you fine people have any overall advice on the best way to communicate customer-specific recommendations between R / Azure Container Services and a web-shop hosted through Azure? I'm thinking that the container can be set up to read data from sources (eg. a database), and write it's recommendations to some cloud storage area that's accessible to the web-app.
Does this sound like a reasonable approach, or should I keep looking for an alternative way to set this up?