IT Logic

I’m Klaus Meffert. I am the CEO of IT Logic GmbH in Idstein, which is located between Wiesbaden and Frankfurt. While working as an IT expert, I studied for my doctorate. I am naturally very interested in AI as a topic because it is technically highly complex, which is very appealing to computer scientists and is also very closely linked to data protection and data security. I've been really involved with this topic since 2017 and still am. I have a blog called “Dr. DSGVO”, which is concerned with those subjects. AI, dealing with highly technical and large amounts of data, digital data protection. That's exactly what we give ourcustomers.

The problem is relatively simple. Perhaps everyone knows a magic square. Four by four, for example. You put the numbers 1 through 16 into the square so that they are all in itat once. And in such a way that all columns, rows and the two diagonals have the same sum. That is relatively complex and a magic hexagon is a bit more complicated. So complicated, that it's not really possible to do just try putting in numbers by trial and error one after the other. It would take forever. Instead, you have to think of more clever ways to solve the puzzle. So, yeah, that was the magic hexagon.

Yes, well, in the past, I tried to approach the problem with Java. It's a compiled language, which runs very quickly. This time I tried the number-1 AI programming language —I used Python. I thought up my own framework program which had an approximate path to the solution. I wasn't exactly sure myself what I should do to solve the complex mathematical aspects of the problem. Then I used two AI mathematical models, which were running on the servers. I used Hetzner servers for that. The second thing I did was to accelerate the program I used by incorporating a just-in-time compilation. And a different AI model also helped mewith that. With that I could make the code much faster. Thirdly, I had another possible approach to the solution in my head, but couldn't really try it out, and wasn't sure how I should go about doing it. The AI mathematical model helped me once again to find possible false starts, and to filter those out from the beginning, so that I could focus in on the actual problem and reach a solution.

Yeah, it must have been two of the GEX44 servers because those are GPU servers — meaning AI servers with Nvidia graphic cards. The model itself is quite affordable to rent and you have lots of options for running AI models. For me, Nvidia is the basic prerequisite for being able to do decent AI programming, because essentially every library runs on Nvidia graphic cards on the Cuda system. There are only a few alternatives. The price is great. The graphic card is good and has 20 gigabytes of VRAM. That's enough to be able to run many AI models, so that is one reason. The second is the low price, of course. So you don't have to consider whether or not you can really rent the server. And the third is the data protection, of course. I trust Hetzner as a German provider 100% with that — German data protection agreement, german location — that is perfect for me.

Yeah, sure. From my point of view, Hetzner speaks the customer's language. I feel very well taken care of. As a provider from Germany, Hetzner offers me maximum data security and AI servers at a great price.