Post cover image
Photo by Kaitlyn Baker on Unsplash

January 31, 2026

How NOT to Become an AI Engineer in 2026.

Avoid these mistakes.

By Marina Wyss

5 min read

AI engineering is one of the best career opportunities right now, and yet I keep seeing people spend years studying with almost nothing to show for it.

If you've been learning AI engineering for a while and still feel stuck in tutorial hell, I can tell you it's not about effort — it's about direction.

Today we're talking about the mistakes that are slowing you down, and more importantly, what to do instead.

What Even Is AI Engineering?

But first, we need to get clear on what an AI Engineer actually does, because I think this is where a lot of the confusion starts.

When someone says "AI Engineering," they're often describing data science or traditional ML engineering, which are actually quite different roles. I have a whole video breaking down the differences but here's what you need to know:

An AI Engineer is NOT primarily someone who trains models from scratch. Instead, they build applications on top of pre-trained foundation models like GPT-5 or Claude. They work on things like prompt engineering, Retrieval-Augmented Generation, fine-tuning, and AI agents.

And companies desperately need people who can do this. There's a genuine shortage of people with these skills, which is why the compensation is so competitive.

So with that definition in mind, let's talk about where people go wrong.

Mistake #1

Mistake number one will upset some of you. Most people in the industry will tell you that to become an AI engineer, you first need to deeply study calculus, linear algebra, and traditional machine learning — calculating derivatives by hand, working through proofs, and spending months on the mathematical details.

I disagree. But not for the reason you might think.

Here's what I mean: You need enough machine learning knowledge to tell the difference between supervised and unsupervised learning, how models are evaluated, what metrics actually mean, and have solid intuition for how neural networks function.

What you don't need is to calculate the chain rule by hand. You don't need to derive backpropagation from scratch. After six years working on production AI/ML systems, I've never once needed those skills.

The problem is that most learning paths bury you in mathematical derivations for months before you ever build anything. And every hour you spend on hand calculations is an hour you're not spending building the skills that will actually get you hired as an AI Engineer.

Remember what we talked about — AI Engineers build applications with pre-trained models. You need conceptual understanding, not derivation skills. Intuition, not calculation.

So if you're six months in and you're still stuck doing math homework, you're optimizing for the wrong role.

But here's where it gets tricky — because even when people understand that AI Engineering is about building with pre-trained models, they make a different mistake that's just as costly.

Mistake #2

Mistake number two trips up even people who understand they're building with pre-trained models. They get so focused on prompts and model performance that they forget that the model is just a little piece of the project. AI Engineers build entire systems.

Your chatbot might work great on your laptop with three test users. But what happens when a million people try to use it simultaneously?

Real AI engineering means thinking about evaluation for each component and the entire system. It means considering performance, security, and cost. Basically, how do you actually get your code running reliably in production?

Remember, you want to become an AI Engineer. You need to build entire systems that function at the level of millions of users if you want to be competitive in this job market.

Mistake #3

You might have heard that you need to build a portfolio to become an AI Engineer. And that's true! Building a solid portfolio is one of the best ways to demonstrate skills before you have professional experience.

But you won't be successful with just any portfolio project. It's really easy to waste time building the wrong things.

If your portfolio consists of toy projects and ChatGPT wrappers, this is the AI Engineering equivalent of the Titanic dataset in data science. It's way too simple and has been done by everyone else you're competing with.

Instead, mimic what we build in the real world. Ideally, build something actually useful that solves real problems for real people.

What does that look like? It means projects that cover every aspect of the AI Engineering lifecycle — all those things we talked about like evaluation, observability, deployment, and security. Your portfolio should show a hiring manager you can do what they need you to do for their company, not just that you can follow a tutorial. If this sounds overwhelming, stick around to the end for some resources to help.

Ok, so at this point you're building real projects with production concerns in mind. You're on the right track. But there's another way people sabotage themselves, and this one is easier to miss.

Mistake #4

Learning AI engineering is overwhelming. There are hundreds of tools and concepts to learn, and the hard part is that the list of things to learn keeps growing! It's like as soon as you've learned one tool, five new ones have come out. You try to keep up but it just feels impossible.

But actually, trying to keep up IS the mistake.

If you're getting distracted with every new tool release or getting hung up on specific frameworks, you're not focusing on what really matters.

Instead of spending six months becoming a LangChain expert, shift your attention to the core underlying concepts. When learning frameworks, the goal should be to develop a mental model of the underlying architecture so that you become a true AI Engineer, not someone who just memorized the API documentation for a particular tool that may be obsolete within a year. The specific tools are just implementation details.

Ok so at this point you've learned the right things and built a solid portfolio. So why are you still not getting interviews?

Mistake #5

It's frustrating when you feel like you've done everything right and you're still not getting job opportunities. You've learned the skills, made the portfolio, and applied to hundreds of jobs on LinkedIn with nothing to show for it.

But the reality is that even if there are a lot of AI engineering jobs right now, there are also a lot of applicants. If you have no experience, you have an uphill battle. It's you versus a few thousand other people applying to that same job on LinkedIn.

So what are you supposed to do?

You might not like to hear this, but the highest leverage activity is networking. Even if it's scary, and you don't know anyone, making direct connections is the best way to increase your chances of being given an opportunity.

This is how I got my first job in the industry, so I know it can work. But you have to do it the right way. That means:

  • Connecting with people at companies you're interested in
  • Engaging thoughtfully with their content
  • Asking for informational interviews
  • Contributing to open source projects where people will notice
  • And building in public so people can see your work

None of this is rocket science. It's just work that most people won't do because it feels awkward or time-consuming. Which means if you do it right, you have a real advantage.

— — —

If you're feeling like you need some support with your AI/ML career, here are some ways I can help: