Quoted By:
In terms of capital requirements, something mathematical and/or computational.
ML specifically lets undergrads publish absolute dogshit at conferences, literally everything you can think of can be a paper, so you could ride that wave I guess.
As far as amateur mathematics, you can probably chip away at some low-barrier-to-entry combinatorics problems, even if it's just exploratory, or document some integer sequences.
In terms of utility, you can be useful as a "citizen scientist" in various field and observational science. Amateur astronomers attend certain conferences that are pro-am. There are various ecological databases to contribute to, paying attention and logging data is somewhat underrated. Most people think that the goal is to identify some new species or something, but dumb shit like soil samples or picking up some haircut moss and checking out what it has been sucking on can give people insight into at least local dynamics.
If you're actually a decent programmer, making high performance open source tools for various research domains is going to be high value add and can certainly net you some citations if that gets you hard.
Biological experimentation at home is getting easier, but I have less than 0 knowledge about what is really possible there, I only know that the cost to tinker at home is dropping (The Odin, for example).
I have to imagine anything experimental beyond that is too costly for an amateur.
Of course, a final option is to get really good at statistics and just check others' work (like Gelman).
