4 min. read
An interview with Piotr Gajda, Moderation and Trust & Safety Team Leader at OLX Poland – on building a tool that moved his team from reacting to threats to seeing them coming.
Piotr, tell us who you are and what your team actually does.
I’m the Moderation and Trust & Safety Team Leader for the Polish market, managing a team of specialists across OLX, OTOMOTO and OTODOM. My team is the first and last line of defence when fraudsters try to target our users – whether by stealing money or personal data. We respond to eliminate threats before they cause real harm.
Beyond fraud and phishing, I’m also responsible for intellectual property – detecting and removing counterfeit goods, and building long-term relationships with brands and companies that need platform-level support in that area.
Fraud doesn’t stand still. How fast do fraudsters actually adapt?
Within hours of us detecting their scheme – sometimes faster. This isn’t an environment where you implement a solution once and move on. I saw this clearly while building our tool. Every change on our side triggered a counter-response on theirs. They’d change domains, modify content, test new attack vectors.
What they do is largely A/B testing – except the goal is to bypass protection.
For teams, this means one thing: tools must be alive. Static rules work for a moment. Real protection is continuous observation, rapid adaptation, and the readiness to accept that what worked yesterday may be obsolete tomorrow. That’s why from the beginning I focused on making our tool flexible – so it could be quickly modified without depending on long deployment cycles.
What triggered you to build this tool in the first place – and what did the first version look like?
OLX already had tools to fight fraud and phishing. I didn’t start from nothing. But there was a moment when I saw that despite existing safeguards, specific campaigns were slipping through the gaps. And I knew that waiting for other solutions was a luxury we couldn’t afford at that moment.
The first version of the tool was very simple. I literally called it “the bag.” Data poured in, but most of the analysis was manual. Someone had to sit down, review everything, draw conclusions and make a decision. I quickly understood that it wasn’t scalable. The real value would only appear when we automated what could be automated – and that became the focus of every iteration that followed.
At what point did the data start revealing things you couldn’t see before?
That was the breakthrough. When the data became good enough and we started connecting it in real time through automation, we suddenly started seeing things that were previously invisible – patterns, connections, repeating behaviours. And crucially, we started seeing them within seconds.
If you can see within seconds, you can respond within seconds. That’s a fundamental change. Instead of analysing an incident after the fact, we could intervene before it had a chance to cause harm. We could see the same fraudster trying again – different profile, different domain, but the same behavioural fingerprint.
That kind of visibility changes not just the speed of response, but the entire philosophy of protection: from reactive to proactive.
What did OLX do that made building this possible?
Several things came together at once. My manager, Piotr Turlej, helped shape the first idea and trusted my vision – that gave me the space to act. Without that trust, the concept stays a concept.
But equally important is that OLX genuinely invests in access to tools. Top AI tools, process automation tools, planning tools – which in larger projects is often underestimated but hugely important. This isn’t an environment where you have an idea and have to fight for basic resources. You have what you need to work with, and you can turn a concept into something real much faster than in many other places.
Where does the tool go from here?
The direction is clear. Protection must become increasingly intelligent and increasingly less dependent on manual intervention. Deeper decision automation, better anomaly detection – and greater collaboration between OLX markets.
Fraudsters don’t operate locally. A campaign that starts in one country can appear in others within days. The tool has the potential to become a protection layer that works across markets – sharing knowledge, patterns and signals in real time. The next version is a tool that doesn’t just react. It acts ahead – before an attack fully develops.
What’s the most important thing you’ve learned about building security tools in the AI era?
AI has changed the rules of the game on both sides. We use new tools to detect better. But fraudsters use exactly the same tools – and they do it very well. They generate convincing content, scale attacks, and personalise phishing in ways that just a few years ago would have required a large team and significant resources.
So I’ve learned humility in the face of this dynamic. With the mindset of “I built it, it works, done” you’ll build a tool that becomes useless faster than you think. Real security is continuous vigilance, continuous evolution – and acceptance that you’re in this for the long haul.
Piotr story is part of our “AI Innovators” series. Across OLX, people are finding new ways to use AI — to create, simplify, learn, and move faster. This series celebrates the colleagues who are sharing their discoveries and inspiring others to experiment along the way.
