Closed epec254 closed 2 years ago
Thanks, this is moving towards a distinct challenge that needs to be called out. Can we perhaps tidy this up just a little to simplify the message? My suggestion would be to focus on the product management aspect of this challenge specifically, rather than the detail of technical activities which we already have in other challenges. I think we are really missing the clear directive that customers want integrated solutions, not steaming piles of components. We have discussed the specifics of implementing that in other challenges but we aren't actually tracking the fundamental driver itself, and that need is not yet widely understood in the ML vendor community.
Updated based on Terry's feedback.
After discussion with @tdcox and @AlmogBaku , proposing a specific challenge around the fact that predictions from ML models are (almost) never used as stand-alone outputs themselves, but rather, these predictions are embedded within other products (apps, reports, etc).
Following how the rest of the guide is written, I have avoiding naming ML assets or using any vendor specific terminology and tried to write that language such that it was focused on ops (e.g., it was relevant no matter the underlying ML approach used).
I believe a complete solution must address these challenges.