The basic and best method is apprenticeship - attach yourself (somehow) to a Master: someone who can do it already. Help them with their work (without pay), and in return they may teach you, advise you, or you may pick up an understanding of how to do what they do.
Read into the subject. Talk or write about what you read and try to get some feedback. Valuable feedback from a competent 'Master' is very, very rare however - it may come seldom and in little scraps, and the apprentice must be alert so as not to miss it.
Don't be too impatient to find a specific problem to work on - allow the problem to find you. Francis Crick proposed the 'gossip test' - that which you gossip about spontaneously, is probably contains a possible problem to work on.
When you are interested in a *problem*, you can usually find some aspect to work-on which you personally can do with your resources of time and effort, and without lavish material resources or manpower.
Publication is a matter of informing people who are genuinely interested in the same problem. This might be done by letter, as in the 17th Century. The internet has solved the problem of making work accessible to those who are interested.
If you are honest/ can earn trust, produce useful work or provide some valuable function, you will be admitted to the 'invisible college' of self-selected people working on a problem.
If you are not trustworthy, lack competence, or are unproductive, then you will not be allowed into the invisible college - because an invisible college is a synergistic group sustained by mutual benefit. If you don't provide benefits to the group, and show no prospect of providing any in the future, then you are merely a parasite and need to be excluded.
The respect of an invisible college is the currency of science - it is the invisible college which evaluates work, and develops and sustains understanding through time.