The same Kubernetes or MLflow sound cool, but for a small project it can be overkill and only complicates life. I once tried to assemble something like this and half of the time was spent not on the AI ​​itself but on getting it all together properly. But on the other hand, it's good that now there is a choice and you can not build everything from scratch. When I looked at how it is done in practice, I came across ai solutions development https://artjoker.net/services/ai-development-services/ and there it shows more real cases. How it is assembled for tasks and not just a list of technologies. And it helps to look at things a little more soberly. Therefore, I would say that the stack is important, but even more important is how you use it and for what tasks.