Everyone talks about the CS degree like it’s either worthless or a golden ticket. Neither is true.

Your degree has real advantages. HR takes you seriously. Employers in Germany and most of Europe still look at it. You build a network of people who’ll be working in the field. You learn the basics: networking, algorithms, how a computer actually works. These are not nothing.

But the degree alone won’t get you hired. The gap between what universities teach and what companies actually need has never been wider. And if you’re applying right now and not hearing back, that gap is probably why.

Here’s what bridges it.

the three skills that actually matter

1. infrastructure

Every application needs to run somewhere. That somewhere is infrastructure. And most CS grads have never touched it.

In a production environment, you’ll typically have a server running Linux, with Kubernetes on top of it, managing your applications. Infrastructure engineering means knowing how to provision and maintain those servers, how to deploy applications on top of them, and how to automate as much of that as possible.

This skill is also one of the hardest for AI to replace. It’s physical, complex, and deeply operational. Learning it puts you in a position that’s genuinely hard to commoditize.

What to learn: cloud providers (AWS is the industry standard), Kubernetes, Linux, automation with tools like Terraform and Ansible.

2. coding (the production kind)

Most CS students think they know how to code because they did data structures at university. That’s not coding for a job.

Production-grade coding means building things companies actually use. It means understanding how software gets developed, tested, and deployed in real teams. It means knowing when to use AI to go faster and when the fundamentals matter.

You don’t need to hyperfocus here. The goal isn’t to compete with engineers who’ve been doing this for 20 years. The goal is to understand how code works well enough to build things and to use AI as a force multiplier on top of that foundation.

3. AI as a force multiplier

This is not machine learning. This is not building language models. This is using AI tooling to do more with the skills you already have.

Every engineer who uses AI well can move faster, write better, and handle more complexity than one who doesn’t. That’s just true right now. An employer doesn’t care if you got a little worse at raw syntax because you use Copilot. They care if you ship.

Learn how to use agentic AI workflows. Get familiar with tools like Claude Code. Understand what model context protocol is and why it matters. This stuff compounds fast.

how to actually learn this (not by binge-watching courses)

The mistake most people make is spending six months on a course and coming out the other side unable to deploy a basic app. Courses have their place, but they’re not enough alone.

The learning model that works: theory and building in balance, leaning more toward building over time.

A realistic daily structure for someone starting out:

  • 30-60 minutes of theory in the morning (course, book, documentation)
  • 30-60 minutes of building in the evening (hands-on project, following along, breaking things)

Even 20-30 minutes a day of actual hands-on work compounds faster than you think.

the concrete roadmap

This is the order that makes sense.

Step 1: Linux foundation. Take the Linux Foundation’s Kubernetes course (Misha Vandenberg’s version is excellent, go during a sale and you’ll pay 10-15 euros). It gives you a strong conceptual base and has a good mix of theory and practice.

Step 2: Install Arch Linux manually. This is not a production skill. It’s a learning tool. Installing Arch manually forces you to understand partitioning, disk tables, and how a Linux system actually gets built from scratch. Do it once with a guide, then reinstall from your own notes.

Step 3: OverTheWire Bandit (up to level 10-15). This is a game that makes you good at the Linux command line. Working in the cloud means working from a terminal most of the time. Get comfortable with it.

Step 4: Learn to Cloud curriculum. Made by GPS, this is a structured set of projects that walk you through real cloud work. Work through the curriculum at your own pace.

Step 5: Build with AI tools. Go to Anthropic’s free courses at anthropic.skilljar.com. Work through the AI agents content. Then build things. Actual projects where you use Claude Code, MCP, subagents. That’s where the leverage starts to show up.

The whole path will take you 6-12 months of consistent daily work. That’s the honest timeline.

the thing nobody wants to hear

You can’t shortcut this by watching more content.

The engineers getting hired aren’t the ones who finished the most courses. They’re the ones who built actual things, broke them, fixed them, and can talk about what they learned. The portfolio and the story that comes from real building is what gets you through the interview.

Open your laptop every day. Even for 20 minutes. Build something. That’s the job.

If you want a free structured path through all of this, the CareerLaunch challenge is on Skool.