I believe that I will be closing this thread since I'm not sure that this is quite what the ML section is for.
The original purpose was to upload logs that had been manually gathered on an offline server.
Although the solution is not perfect as I still haven't found a way to have the data parsed through winlogbeat pipelines, I did manage to send large series of ndjson data through filebeat, which I consider the (partial) solution for the time being.
I'm also exploring ways to achieve this via logstash in this thread, in case anyone's interested.