1. You started here as a working student about three years ago and have became a permanent member of the team in the last two years - what was the biggest change between studying at university and working?

Whereas during my studies I sometimes got stuck on something on my own, I appreciate that at this job I'm part of a great team where we support each other and look for solutions together. I also get to take on more responsibility and see for myself how my work adds value to projects, whereas during my studies it often remained only theoretical. That's why the working student job was also a very good experience for me.

2. What originally drew you to NLP development—and what keeps you enthusiastic about it today?

I've always found language(s) interesting, and when I discovered the exciting things you can program for linguistic applications, I was hooked. Now I'm excited that NLP is becoming increasingly relevant for various digitization and automation projects.

3. What challenges do you encounter most frequently in projects—and which of them do you unexpectedly enjoy?

One challenge is often proactive communication, in which all project partners are brought on board. Interface connections for customer systems can also be challenging; it's sometimes like a little scavenger hunt for specifications and settings that still need to be changed, but when it works, it's also very rewarding.

4. Is there a project or use case where you thought, “Wow, this does not only work technically, but it also helps real people”?

Yes, our roadside assistance voicebots immediately come to mind. It's great to know that we are supporting people who need urgent help, and that motivates me to keep at it and continue looking for ways to optimize the service.

**5. How do you deal with it when a model doesn't perform as expected—debugging, magic, or patience?

In general, when something doesn't work, it often helps me to get a second pair of eyes or ears to brainstorm, play a quick game of table football, or to simply come back to the problem the next day with a fresh mind.

6. What advice would you give to companies that believe NLP works “just like that” – without data, training, or context?

That would be nice, but effective application requires thorough preparation and regular maintenance. Every company has specific processes, its own vocabulary, and very individual requirements. Therefore, it is more effecitve in the long run to define all of this and work out the needs in the beginning. This ensures that the solution works exactly right instead of being generic.

7. As an NLP expert, what can you talk about at length, even if no one has asked?

Embeddings! These are vectors in high-dimensional space that represent semantic content and can be used to calculate word similarities, for example. I think it's cool that language can be represented mathematically in this way.

8. Is there an absurd or nerdy habit that clearly identifies you as an NLP person?

Maybe not necessarily as an NLP person, but definitely as a language nerd: In German, you can string nouns together so beautifully, and I always stumble when they are written separately, e.g., die Erdbeer Marmelade (the strawberry jam) or der Obst Salat (the fruit salad). Long live compound words!