weighted-random | Select randomly from a list of weighted values

Let's say I have a list of objects (in Python) that looks something like this (contains an identifier and a ranking/weighting):

I have a bunch of names from the web (first name, last name, of people in different countries). Some of the countries have statistics on how many people have each last name, as shown in some places like here.

Well, that Japanese surname list only lists the top 100. I have other lists like for Vietnamese listing the top 20, and other lists the top 50 or 1000 even in some places. But I have real name lists that are up to the 1000+ count. So I might have 2000 Japanese surnames, with only 100 that have listed the actual count of people with that surname.

What I would like to do is built a "faker" sort of library, that generates realistic names based on these statistics. I know how to pick a random element from a weighted array in JavaScript, so once the "weights" (number of people with that name) are included for each name, it is just a matter of plugging it into that algorithm.

My question is, how can I "complete the curve" on the names that don't have a weight on them? That is, say we have an exponential-like curve sort of, from the 20 or 100 names that have weights on them. I would then like to randomly pick names from the remaining unweighted list, and give them a value that places them somewhat realistically in the remaining tail of the curve. How can that be done?

For example, here is a list of Vietnamese names with weights:

Using randint() how do I give lower values a higher weight (higher chance to be picked)?

I have the following code:

Is there a way to do a general, performant groupby-operation that does not rely on pd.groupby?