1 comment

[ 19.0 ms ] story [ 426 ms ] thread
Yep.

If you are not curating your own training data, your AI project will not succeed. A feature store is necessary, but any particular algorithm is optional.

I would say though to give up on "MLOps" and just have "OpsOps"; that is, build the system that breaks a task into decisions and make it possible for people to do it 100% manually, get the AI system to make decisions when it can.