Starting my first VPN project – how to design a scalable backend?

[Aleks-7-7-7]

Hi, sorry if I’m asking this in the wrong place — I’m not sure where exactly to post questions like this.

I’m a self-taught junior developer, and I’ve started working on my first pet project — VPN service managed through a Telegram bot. Eventually I’d like to monetize it, but for now I’m focused on building something that works and learning along the way.
I’m a bit lost on what tech stack to choose and how to describe the backend architecture clearly, especially if I’ll need help from freelancers. Here's what I want to build:

1. telegram bot that receives payments and returns a personalized subscription link.
2. backend server that accepts a GET request (subscription link) and responds with a base64-encoded VLESS + REALITY configuration for the chosen location.
3. Xray-core-based infrastructure, where clients connect to one or more inbound servers. These inbound nodes should distribute load across multiple outbound servers, possibly in different regions. In the long term, I’d like to scale to 10,000 users.

My questions:

• What’s a reasonable inbound-to-outbound ratio to start with?
• What’s the best way to handle load balancing? Should I rely on Xray's built-in routing + balancer? Or is it better to use external tools like NGINX or HAProxy?
• Are there any existing open-source solutions or APIs that can issue user tokens, track subscription time, and serve user-specific VLESS configs?
• Any practical advice for scaling to a large number of users? Like kernel/network settings, port reuse, CPU/RAM planning, monitoring, failover, etc?

Thanks in advance for any help or pointers.