I just was wondering, how https://phantombuster.com does all those automations on a scale? I mean, I can surely write some code that will work for one user(may be just for myself), but how do I scale it up for millions of users? how does it work on millions of LinkedIn/Youtube/Twitter etc user sessions?
Can anybody give me a hint?

A lot of ways.

Caching on all three levels, server side caching of data in nosql databases.
Web accelerating caching at the level of varnish/nginx
Client side caching.

Load balancers to scale horizontally.

Event based system and architecture around kafka for example. With multiple workers processing the river of tasks.

Probably many other ways.

The phatom thing looks like built around event flowing system.

With multiple workers processing the tasks.

It can be even made auto scaling, depending on current load.

User sends event to stream.
One of their multiple workers processes the new task in the queue
Users listens and gets the answer.

https://www.gentlydownthe.stream/
Here super friendly explanation for how kafka works

@darkwind whoa! awesome man! thanks a lot!

15 years ago I worked for a website that automatically scaled the amount of servers ( back + frontend + cache/edge ) based on the amount of traffic on the website with a min of 70 servers running at all times. ( And while I was there I saw them scale up to 1500 once ). But that 15 years ago, technology has come really far since then, you can only imagine what they can do with autoscaling these days

As for sessions : https://docs.aws.amazon.com/elastic...
( Other vendors are available too, I just know AWS the best )