Experience with Nvidia/Cuda/cudnn, Tensorflow, and vscode?

Anyone using Nvidia for data science?
I have an Nvidia RTX A4500 and have been trying different distros, distroboxes, VMs, etc. trying to find something stable/reliable to do my work on.
I was previously using Pop!_OS, but it gradually degraded, with long pauses that got longer, then Nvidia driver errors.

Currently, I have an Arch distrobox running on Universal Blue which I access in vscode as a remote container. While it works, there seem to be a lot of moving parts, and some things are “mysterious”, like being sure the container and system are being properly backed up.

So I watched some EndeavourOS videos, and it looks great! Since my Arch distrobox works, I thought maybe a bare metal install of EndeavourOS would be the way to go.

One small caveat (perhaps) is that my /home is encrypted btrfs and on a separate drive.

Thanks for any experiences you can share!