Member-only story
How I’d Learn Machine Learning in 2026
If I could start over
I’m a Senior Applied Scientist at Twitch and have been deploying production ML systems for the past seven years. If I had to start over from zero today, in 2026, I would do almost nothing the way I did the first time.
And honestly, I’d ignore most of the ML career advice out there right now, because a lot of it is from people teaching the path that got them hired (in 2019). That path won’t work today. I know this since I’ve coached over 200 people into the field, and I’ve seen how much things have changed.
So I’m going to walk you through what I’d actually do instead, step by step. And if you’ve been grinding through a calculus textbook for the past three months, this might make you rethink your plan.
Intuition > Math
But of course, I did the exact same thing. I spent months on math before I ever touched a model. I read math books cover to cover, deriving the chain rule by hand, the whole shebang. And it worked…eventually.
It took years longer than it needed to, and the weird part is, all that rigor didn’t actually give me great intuition for why models behave the way they do in practice. I could explain the math behind gradient descent, but when a model was struggling in production I was still kind of…
