[barefeg]

The answers so far tend to be either: yes, it’s very easy. Just follow a boot camp or books, transition at work by working close to a DS/ML team, etc., or no, you need lots of background knowledge, track record of publications, etc.

I think the problem seems to be that there’s not a clear need for engineers with lower levels of expertise in some of the fields; that is people with less than an MS degree in that specific ML sub field.

If the field transcends for long enough, and urgent demand becomes so mainstream, then I imagine hiring managers will have to invest more resources in hiring people with less experience and training them in the job. Similar to how developers are hired fresh out of short boot camps these days.

Therefore the easiest way to transition would be by acquiring practical an theoretical knowledge using the strategies given in the other answers and then applying to ML teams where there’s enough demand for them to want to train you in the job. Of course that’s easier said than done. It might be interesting to hear some thoughts on whether this is already happening in certain fields.