I studied machine learning when I was in university, a couple of years ago. I used it for my master thesis (decision trees, ensembles, SVM, and word embedding mostly) and for other projects either personal and academics (genetic algorithms, Q-learning).
Then I studied neural networks by myself, from the pure theory of the perceptron, backpropagation, gradient descent, etc. (with Bengio's deep learning book) to the various ramifications of ConvNets, DeConvNets, auto-encoders, LSTM, deep reinforcement learning, and other various papers. I took a several months break and there are already new types of networks that I don't know, so hard to keep up with the state-of-the-art.
However, my knowledge is purely theoretical. I want to embed machine learning in my future career, but getting a job in the field requires practical expertise, which I lack.
I am working full time now as a software engineer. So, what do you think may be a possible path to follow in order to gain expertise? Given that I have a strong mathematical background and more than 10 years of coding on my back, is there any resource on practical rules of thumbs, suggestions, real case studies, etc., which I can invest time onto? I mean something beyond the vanilla "introduction to machine learning", something more real and concrete.
Do you think that attempting Kaggle challenges from scratch would be stimulating? Or too confusing? Do you think delivering a personal project (i.e. not a byproduct of my current job) in which I employ one such method to solve a potentially real problem would be enough to be relevant in my cv?