I've been training neural networks since before it was cool (circa 2016), and have been fortunate to contribute to a number of incredible innovations in AI through my work at Ezra:
- The first FDA-cleared Prostate AI in the world
- The first radiology report to lay-term translation AI
- The world's first AI-powered 30-minute full body MRI
I sometimes get asked how one can go about becoming a machine learning engineer, so I've put together a list of my favorite resources. I recommend you go through them in the exact order below.
- Neural Networks: Zero to Hero by Andrej Karpathy. This series will teach you how to build a neural network from scratch, including forward feed, backprop (including calculating the gradients manually). It's very useful for understanding the internals of a neural network.
- Practical Deep Learning by Jeremy Howard. Once you understand how a neural network works, it's time to get some actual practice building them (without worrying about the internals). I have found it's most useful to go through this once you've done Andrej's course above, as you'll be able to follow along much better.
- Read Practical Deep Learning for Coders. Jeremy Howard wrote this book before creating the Fast.ai Course (above). I recommend you do the course first, then read through the book and / or use it for reference.
- Read Deep Learning with Python by François Chollet. Even though the book uses examples written using the Keras library (I prefer PyTorch), it does a pretty good job drilling the key concepts into your brain.
Going through the stuff above will give you a very sound foundation for becoming a machine learning engineer. From there, the only way to learn is through practice. Find an area you're interested in (for me, it's healthcare imaging data), find a Kaggle project in that area, and try to break into the top 10 for that project's leaderboard (FYI it's going to be pretty hard).