If you're switching careers and moving into data, building a portfolio is essential for making the first step. You need projects to get hired, but you haven't been hired in the data field yet…So where do the projects come from?

That question used to be harder to answer. In 2026, with AI tools that can help you clean datasets, generate code, and structure analysis, it’s  much easier to turn ideas into real projects. However, that has raised a different, more important question: If AI can produce the output, what exactly is your portfolio supposed to prove?

The answer to that is worth understanding before you build anything.

What Greek employers are actually looking for now

A few years ago, a portfolio for a career changer needed to demonstrate technical competence above everything else.

Can this person write SQL? Can they build a dashboard? Can they work with real data?

Those questions haven't gone away, and in the Greek market, where many hiring managers are still relatively new to evaluating data roles, they still carry real weight. A clean, functional project that shows you can handle a dataset end-to-end will get you further here than it might in markets where technical screening is more automated.

On the other hand,even at the local level, something has changed. As more Greek companies invest in data teams, and as AI tools become a standard part of the workflow,  the conversation in interviews has started to move. Hiring managers are less impressed by the output itself and more interested in the decisions behind it. 

Did this person understand the business problem before building anything? Can they explain their thinking clearly to someone non-technical? Do they know when an analysis is good enough to act on, and when it isn't?

That last question matters more than most candidates realise. Greece's data market is still maturing, which means many of the roles available right now sit at the intersection of technical execution and business communication. Pure technical ability will get you noticed. The ability to translate analysis into something a business can actually use is what gets you hired, and what makes you genuinely useful once you're in the role.

For career switchers, this is the opening. The domain knowledge you carry from a previous career, understanding how decisions actually get made, what questions matter to a business, where the real uncertainty lies , is something that takes time to develop. A portfolio that reflects that context doesn't just show technical ability, it shows professional maturity. 

AI is a tool, not a shortcut

However, there's a version of portfolio-building that uses AI the wrong way.  For example, you may be able to prompt an LLM to generate a full analysis, clean it up slightly, and present it as your own work. It might look credible on the surface, but it tends to fall apart the moment someone asks you about it in an interview.

The better approach is to use AI the way a professional would: as a collaborator that speeds up the execution, not a replacement for the thinking. Use it to help debug code you've written, to suggest approaches you hadn't considered, or to explain a statistical concept you're unsure about. Keep your hands on the problem throughout.

The projects that stand out aren't the ones that look the most polished. They're the ones where it's obvious a human was genuinely engaged with the problem-  where the documentation shows real thinking, the README acknowledges what didn't work, and the conclusions are honest about the limitations of the analysis.

What to actually build

Now that you understand what employers are looking for and how to use AI effectively, the real question becomes: where should you start? 

Pick problems that connect to your previous career. This is the most underused advantage career changers have. If you spent, for example,  five years in marketing,you understand the real questions people in this industry are trying to answer. That domain knowledge is genuinely valuable, and a portfolio project that reflects it will be far more memorable  than another generic sales dataset analysis that hundreds of other candidates have already added to their portfolios.

Also, you don't need many projects. Two or three that are well documented and show different types of thinking will serve you better than ten rushed ones. For each project, the most important thing isn't the complexity of the technique,  it's the clarity of the story.

What was the question? What did the data say? What would you recommend, and what are the limits of that recommendation?

GitHub is where the work lives

Host everything on GitHub and treat your READMEs as part of the portfolio itself, not an afterthought. A well-written README that explains the business context, walks through the approach, and honestly notes the limitations tells an employer more than the code does. It shows you can communicate, which, in a data role, is half the job.

If you used AI tools during the project, you don't need to hide that. Mentioning how you used them,  and more importantly, how you verified and built on what they produced, actually demonstrates exactly the kind of critical relationship with AI that employers are looking for in 2026.

Start before you feel ready

The most common mistake career changers make with portfolios is waiting until they feel technically confident enough to begin. That point rarely arrives on its own. The process of building the portfolio is part of how confidence develops.

In short: Pick a dataset related to something you already know. Frame a real question. Work through it, document the thinking, and put it on GitHub. 

Yes, at first it won't be perfect. It doesn't need to be. What it needs to be is honest, clear, and yours.

That's what gets the conversation started.