WBJ: Recently everybody said you had to structure your data in order to have the analytics or any kind of AI on top of your data. But now more and more companies are trying to introduce and analyze unstructured data. How exactly are you able to accomplish this? What are you doing with your unstructured data?

Piotr Stefaniak: Twenty years ago, when I studied computer science, everyone was thinking about how to transform unstructured data into structured data. Even when I was in the UK, I was in a special course with people who invented the worldwide web. They were trying to create a so-called Semantic Web, which meant adding structure, metadata, but something even above metadata; creating a language understandable for machines. Because that was the biggest challenge back then – machines were becoming increasingly powerful, but the data was still unstructured.

Now what is happening? Machines have gotten even more powerful and things that we were learning about 20 years ago actually came true. Machines can analyze the data, learn from it and then basically act on it, etc. Now we all talk about GenAI, but actually, it's even more complicated than we think. Behind GenAI, there are thousands of processors and a great deal of compute time. The neural networks behind it take a lot of time and power to train.

So when I was a student, I was developing neural networks. I was in an AI course, and we were developing neural networks to recognize pictures. For example, there was a picture, and the aim was to recognize that picture as digits. There was training data and test data. Even for such a simple task, of recognizing pictures, it took a long time to produce something meaningful.

For years, AI was used for things like image recognition. With automatic passport control at airports, we show our passport, we place our fingers on a scanner – AI is there. The machine takes a picture of us and compares it to the picture in the passport. Or imagine we have a terrorist we are looking for and there are cameras everywhere. Those cameras are not only for recording. At many airports, cameras are used for online verification. The AI is looking for suspicious people. It sounds scary, but if you're not a terrorist, then you're fine. It's scary for bad people.

So that was before GenAI. With GenAI, as the name says, AI can generate something new – pictures, videos, texts. It uses enormous data and enormous computing power and energy to train. I recently read an article that Trump wants to build more nuclear power plants just to provide cheap power for the tech industry in the U.S.

Relativity originated in the U.S. Our founder is an American with Polish roots, Andrew Sieja. He was a consultant in a legal industry and realized there was a lot of manual work that could be automated. That was the idea. And since then we have created a large company with a very meaningful product that standardizes the whole market of e-discovery.

We have been using AI and machine learning for a while. But about a year or two ago, we implemented GenAI, which simplifies things even further. Imagine a dispute between two companies or a fraud case. Companies today have thousands of employees. They all produce emails, documents, spreadsheets, short messages. The data is unstructured and spread across time, making it very difficult to understand what happened and when. Our solution can take terabytes – even petabytes – of data. Instead of hundreds of humans printing and reading documents for months, you put everything in the system, and a handful of consultants can analyze it in days. With GenAI, instead of a handful of consultants, one person can get the answer even in a one or two hours.

So they don't really need technical skills? Do they use GenAI to interface with the system? Do they ask questions in the natural language of the data that they analyze?

Yes, that's something that was introduced with GenAI.

So the solution is like a component that you put on top of data storage, regardless of where it is stored?

Yes. We interface with all kinds of systems, like Microsoft Outlook, Office 365. You export data to our cloud, the system analyzes it and structures it. If you use the GenAI feature called AIR, it learns based on the data, and then you can talk to the system in natural language.

So is this only used ad-hoc, like when something happens, or is this used continuously?

The main case is to analyze data when something happens. That’s the most important use case.

Do you have a growing market right now?

Yes, the market has been growing at a good pace. We are the market leader. Our primary markets are the U.S., the U.K., Australia, and Canada. Markets like Germany and Switzerland are growing. In Poland, we do not actively sell, but we do have users. Our sales force does not focus on Poland.

But you have a Polish office, right?

Yes. One-third of the whole company is in Poland. Globally we have about 1,700 employees, and around 600 are in Poland. When I joined six years ago, we had around 60people, so we multiplied by ten.

Are there only engineers in Poland?

The majority are focused on the product: engineers, security, product management – around 70%. We also have product support and some corporate functions. Engineering is the most essential part. Our GenAI solution is developed here in Poland. Our main office is in Kraków, with employees working remotely all over Poland. 

So why did Relativity choose Poland? Because of our talent pool?

The decision was made before I joined, but yes. One reason is Andrew Sieja’s Polish roots. Another is the talent base: Poland has 40 million people, more than Romania or Czechia, so there is a proportionally larger tech talent pool. Also, there are strong universities and the presence of other tech companies helps attract talent.

In a time when big tech companies are letting people go, your company seems to be growing.

Yes. We have been growing globally in revenue and employees, and in Poland, growth is tremendous – around 100 new people each year. Layoffs in the market happen in cycles. Companies lay off, and a few months later, they hire again. The U.S. tech market adjusts very quickly. I don’t think the layoffs are permanent. It’s temporary. That’s how the economy works.