It depends on a couple different variables:

- Do you want to focus on a specific area with decades of history (e.g. computer vision), or do you want to be a jack-of-all-trades person that works with images, text, time-series data, and so forth?

- Do you want to work at an early-stage startup outside of the Bay Area, which might be less picky with hiring... or do you want to work at OpenAI or Google Brain?

- Do you want to work in a team where you implement the ideas of better-educated, senior colleagues... or do you want to be the one in a leadership role?

The "minimal" path probably involves:

- A Galvanize (or similar) bootcamp to meet peers and mentors.

- Books and Coursera courses to fill in everything the bootcamp doesn't teach you.

- Winning some Kaggle competitions to show that you can build fairly complex models that have high accuracy on real-world data.

The "maximal" path probably involves:

- An undergraduate degree in Computer Science, Applied Mathematics, or Statistics.

- A PhD in one of the many subfields in machine learning.

- Several industry internships in between.

Neither the minimal or maximal path guarantee success, of course. A lot of that depends on your aptitude, previous experience, ability to network, and the current job market.

I actually just did this. Courses online such as Andrew Ng's Deep Learning course on Coursera and Standford course papers online are where I started. Then began working through tensor flow tutorials. Then looking up papers on modern neural net architectures. Took a matter or a few months to be productive. In the end, I realized that although neural nets are very fascinating, working on training them is tedious and not as fun as working on product code. I'd rather use the neural nets in a product, much like a sensor, than work on actually training the net.

Same here. I too enjoy writing functional code more than I do training neural networks. Most of the classes focus on the training part, but most of the work in practice is running people down to give you data, and cleaning it.

Even sharing datasets is difficult, and I recently wrote a bit about it. https://blog.helmspoint.com/posts/2017/09/05/sharing-machine...

But that doesn't even cover versioning, licensing, permanence, and provenance. The tools available for machine learning training is still nascent, so the experience isn't great.

The absolute fastest way would be to do a Machine Learning Bootcamp like Galvanize or Metis.

Source: I did not know anything about how to use Machine Learning in my code just 4 months ago, and now I have just started as an AI Engineer.

Awesome, Who are you working for ? And did you take the complete Galvanize course ? Would the part time course be any useful ?

Thanks for the valuable comments guys,