How to become an AI engineer
Not long ago, everyone wanted to be a data scientist. Now they want to be an AI engineer.
It’s a fast-growing, exciting, and lucrative role. At its simplest level, an AI engineer builds software where models are a core part of the system. Usually LLMs.
To do this well requires a daunting list of skills: software engineering, machine learning, data engineering, and more. But you don’t need to master them all to get started.
The best way to learn is the same as it’s always been: build things. It’s the fastest way to discover what works and what doesn’t. And it gives you the confidence to talk about things you’ve actually done instead of things you only want to do.
I can’t tell you what to build, but your projects must map to specific outcomes. Just as "Learn French" is a terrible goal, “Learn AI” is too. It needs to be specific, more like "learn to order dinner in French."
And to do this, I want you to lean into AI assistance. You must become at one with the machine.
Just as kids in the 90s had an advantage by growing up computer-native while their parents pecked at keyboards as one-finger typists, the next generation of great engineers will be AI-native. A single programmer working symbiotically with AI can now add more value in a day than entire legacy teams could in a week.
This symbiosis expands your reach too, supercharging what you can build. Mastered backend logic but hate frontend frameworks like React? AI bridges that gap. You are no longer constrained by syntax you‘ve memorised; you are only constrained by your ability to specify what you want.
And to do that well, you’ll need the fundamentals. As I’ve written elsewhere, you don't learn the fundamentals so you can write the code instead of the machine. You learn them so you be exponentially better with the machine. You’ll need to spend time learning about system design, traditional machine learning algorithms, and the underlying mechanics of how LLMs actually work.
Which brings me to a meta point: if you want to succeed, you need a genuine interest in the field.
AI is the fastest-moving industry in the world, and staying current requires real effort. But if you step up, you’ll quickly find yourself ahead of most people, who are content to merely get swept away.
And stepping up requires you to actually do things. It’s so easy to want something and do nothing about it. A tactic that still works for me is to write a list of specific goals and things I want to build/learn, and then… do it. It doesn’t have to be harder than that.
You should notice that personal projects, AI-assistance, and mastering the fundamentals reinforce each other. You’ll often be doing all three at once.
And doing this builds something that can't be taught: intuition. The ability to understand something instinctively, without conscious reasoning. It’s what tells you where an LLM is suitable and where its "spiky capabilities" will falter. It’s instantaneous and always on and is a superpower.
If you have a genuine interest in the field and you execute on those three habits, you’ll quickly become a credible AI engineer.
You might be thinking: is that it? I hope you are.