Small business server for Machine Learning & play

Hi folks,

I'm running a small, 100% remote team that does quite a lot of things with data. As we have more and more topics related to ML and our colleagues don't always have a powerful enough PC with them to run the models locally, I thought about building an Unraid server (yeah, we could use a cloud service, but where's the fun in that?) and putting it in the office, so that every team member can have their own (Windows or Linux) VM workspace they could access on demand, as well as a single Windows VM for after-hours play. It'd also probably take on some of the duties from our Synology NAS: backing up the team's PCs, sharing files and running some Docker containers.

So far, the core spec I came up with is:

- CPU: Ryzen 9 7950X

- Mobo: ASUS Prime X670E-PRO WiFi

- RAM: 4x32GB Kingston Fury DDR5 6000MHz

- GPU: ASUS TUF RTX 4090 OC

- SSDs: 2x1TB Samsung 980 (for cache), 1x1TB Samsung 980 Pro (dedicated for the gaming VM)

- HDDs: 3x4TB Seagate IronWolf (I don't expect we'll have that much data that we'll need to store locally)

Now, as I don't have any experience with Unraid (or any physical servers really), I have a few Qs I hoped you could help me with:

1. Do you see any potential hardware compatibility issues?
2. Would such setup (with few VMs having the same GPU assigned, but of course with only one of them being on at the same time) really work and make sense? 
3. Is it possible to have one SSD dedicated only to one specific VM? Would it make sense or not really?
4. What'd be the best way to ensure full encryption of all disks? Using Encrypted ZFS?
5. Is there anything else that I'd need to be aware of?

Thank you very much for your support! 🙏