You can try some kaggle competitions. Try to build a dataset that others could use. Try and implement research papers. Contribute to open source. Start a company. Participate in research projects. Get a deep learning focused job. There’s so much to do.
Thank you for writing this article. This paper came to mind when I was thinking of hybrid approaches. https://www.cs.cmu.edu/~bishan/papers/kblstm_acl2017.pdf
Prof. Mitchell gave a wonderful lecture at my company couple of years ago and I’ve been fascinated by hybrid approaches since then.