depends, do you really enjoy machine learning?

I think you're not asking the right questions.

What you should be really looking for is what suits you more, not what guarantees a better career for you. At least from my perspective.

Trying the thing out never hurts

Ennui is a terrible thing. The monotony in drudgery sure takes away the rapture and alacrity away from life. But the ones who falter and act on caprice increase disarray in their lives. I intend no malevolence, but I fear I have to dissuade you from taking such a decision. There are copious opportunities and I urge you to explore for what is suitable or expedient for you.

Probably building a portfolio out of cool projects using tech that's relevant in the industry will be much more beneficial to you than either CP or ML contests.

If you're yet to graduate, I would personally recommend making interesting (non-trivial) projects on your own — this shows your attitude towards ownership and your potential to innovate and bring an idea to completion, as well as a thorough understanding of how things work, in terms of both high level architecture/design/concepts as well as low level implementation details.

(The following assumes you want a data science job)

But to do that, you need practical knowledge in those subjects, so learning basics and standard stuff from Kaggle can help you develop the right approach towards solving data science problems, just like doing novel algorithm/data-structure-heavy codeforces problems can help you develop the right approach towards designing single-threaded algorithms that are asymptotically efficient. The easier (but more frowned-upon) way to make projects is to look for a problem where you can apply your skills. You can do this in multiple ways, for example, you can go through some datasets on Kaggle and see how they can be used. You should ideally start learning right away by looking at other people's notebooks — don't focus only on the specific models used, but more on the methodology — data cleaning, feature engineering, how they manipulate data and interpret results, the statistical analysis and so on. Do a few such projects and if you have the right approach (systematic + trying to learn from everything you try), you'll find that you're developing intuition about how data works and you'll also have things to put on your resume as a byproduct.

Focusing on learning is very important — you might have fancy projects on your resume, but I have noticed that when people are asked about the details, it often becomes painfully clear that either they don't have an understanding of why they did certain things (sometimes better solutions exist and they're not obvious, but mostly people just think that this is the most standard way of doing things, let's do this), or they straight up lied or left out parts of the truth on their resume ("follow-along" projects are not your own projects in my opinion). If you are being interviewed by a good interviewer, they will find this issue and it will damage your prospects of getting hired.

I'm not sure whether this is feasible for an average student (sorry if this underestimates/overestimates your skills, because I only know that you're yet to graduate and nothing more) in 2-3 months if you have classes and all to attend, unless you work overtime and try to learn at an above-average pace.