Thread by @neptune_ai

“#MLOps standard industry best practices” don’t apply to most #ML teams’ reality.

Why?

Those who write and share best practices are doing ML at a hyper scale.

Those who read and re-share them are doing ML at a reasonable scale.

Companies like Google, Netflix, Uber, and Airbnb are doing an awesome job for the community by sharing their blogs, white papers, and open-sourcing their tools.

But whatever they do, it is shaped (and biased) by THEIR MLOps problems.

Most companies don’t have their problems.

They would love to have their problems, but they don’t.

They operate on a smaller scale & have different (& other) challenges.

And they are the biggest part of the ML industry.

They want to know what’s the best way to do MLOps at their scale, with their resources & limitations

And since there is not much out there specifically for RS companies, they read (& share) whatever is out there.

And the vicious cycle continues.

The moral of the story is this.

Think about
-your problem
-your use case
-your limitations
-your needs

Solve YOUR MLOps problems

Don’t do something just because it is a “standard industry practice”.

Btw, here are some places where you can find good stuff about reasonable scale MLOps:
- @jacopotagliabue, @xLaszlo, @aesroka, @ElectricWeegie on linkedin
mlops.community

- Mlops discord buff.ly/3VE08Gg
- MLOps reddit buff.ly/3WLkTk7
- MLOps blog (and our guide to Reasonable scale MLOps) buff.ly/2Z5bA2J
- MLOps live buff.ly/3TmuKtC