One year professor in Bamberg and BamNLP

9 minute read

Introduction

I am starting to write this post around christmas 2024, at the end of the first year in which I started to work at the University of Bamberg, where I got a full professor position and am heading the research group on “Fundamentals of Natural Language Processing”, to which we refer colloquially as “BamNLP”. It took however quite a while to find the time to finish it.

With this post, I’d like to reflect a bit on the first year and what I think went well and where I could have done things differently. Perhaps my observations are helpful for some people when they find themselves in a similar situation. But I mostly write this to understand better what happened in the last months. I’ll structure this post in the form of decisions I needed to make, and I’ll try to make transparent what I considered when making the decision.

I started in Bamberg in March 2024 and it is my first “Full” professor position. Before coming here, I have been on a tenured position in Stuttgart, at the IMS, a pretty large institute which covers a lot of areas of natural language processing and manages its own programs on NLP for both a bachelor and a master’s degree. The classes are also popular amongst students of computer science programs.

I’ve been happy there, and was very fortunate to be able to work with some very talented Ph.D. students who I funded either through third party grants I applied for or who came with their own funding to work with me at the institute.

Motivation to Move to Bamberg

The question at the time I received the offer from Bamberg was of course, for me, if I should move. Stuttgart has this quite popular program, I was able to work with great Ph.D. students (and Master’s and Bachelor’s students). The environment at the IMS is great, with three full professors and a set of other research groups.

I decided to move for a couple of reasons:

  • I had a tenured position, but only that – I did not have a yearly budget by the university I was responsible for, I did not have positions I could fill, funded by the university. Everything I wanted to do, I needed to ask other professors in the institute for money or I needed to apply for third party funding. While that practically worked fine, it did not really give me a feeling of absolute independence and freedom to do what I want to do.
  • I needed to embed myself in the teaching program – I could not freely decide what to teach, because the classes and the structure was somewhat settled.
  • I felt comparably invisible to other professors in other institutes or more central administrations. That made it difficult to start collaborations across faculties and disciplines. That was mostly a university-internal issue, and not an issue with collaborations outside of Stuttgart, because from outside, I think it was not really transparent what my role was.
  • The salary was not as good as it could be. Minor point, but also one.

There were also reasons which made Stuttgart very attractive to stay – my life centered (and still centers) around that city, my wife lives here, we have our main appartment here. Of course going away is also a challenge for life outside of work.

In short, I hoped to get:

  • A starting package to fund the start of a new group.
  • A yearly budget to do research with.
  • Positions I could hire people on, with university/state money.
  • A secretary and technical staff.
  • A better salary.

I got all of that, but setting this up, building a group on natural language processing essentially from scratch was and is challenging.

That’s all for the intro – just that you know what my starting point was. Now I’ll discuss the various decisions I needed to make. (and I might extend this post in the future)

“Big” city vs. comparably small town

I moved from Cologne to Stuttgart, then we lived in Stuttgart for a while now my private and work life moves (slowly) to Bamberg. What’s good about working at a university in a big city and in a smaller town?

Bigger cities contribute to a better visibility of a university; and it might be easier to find group members who want to live in the bigger city. That was at least my expectation, but practically, I did not experience these challenges. I think, if one has an interesting research profile to offer, it will develop into something visible independent of the city.

I overestimated the importance of a “known” big city to work in, I think.

Stay in established environment or start something new

An established, visible, and productive environment is very valuable. One can have discussions about research every day, which improves the quality of work. One can come up with new collaborations without the hassle of traveling to another place.

It is, however, sometimes also difficult to implement changes, particularly without being “at the top of the hierarchy”. This was not really a challenge for me, because I liked nearly everything in Stuttgart, just sometimes I felt that discussions how to change some procedures were dominated by what people are used to.

This is of course not an issue in a new environment. In Bamberg, I can implement structures in group members, the research culture, and the lecture content as I like it to be. The challenge for me is, however, that not everything can be easily mapped from “I’d like it to be different” to “This is how I’d like it to be.”.

I do not really have a recommendation for this, but this is probably also not really a problem. Being aware of this situation already helps a lot.

Stay in the private environment or move away

Leaving the center of life in Stuttgart was the biggest mistake, from some perspective. I do not like Stuttgart – it is an overly conservative city with a car-centric layout, in which it is possible but not easy to find things I like. For me, it is very important to be able to use my bicycle as the main means of transportation; and that is nearly impossible in Stuttgart. On the other side, own decisions influence other people, and that is a huge challenge. I do not really have a solution for this situation, and I believe that this is the biggest part of the decision to move that can be considered a mistake.

What is good for me is the city of Bamberg – I can cycle everywhere without having cars honk at me or drivers shouting at me. The infrastructure is not particularly good, but so much better than in Stuttgart. In Stuttgart, cars were particularly aggressive from my experience. I enjoy cycling in Bamberg a lot. In addition, I found hobbies that I couldn’t do in Stuttgart; and this aspect made my life considerably better.

Quick start vs. slow start

Some people who start a new group take their time to do this. I think this is a good decision. Understanding the environment, the processes in the administration, the study regulations, the Ph.D. regulations, the processes to buy things for the group – that all takes time.

I was, however, in the fortunate situation that I could bring considerable amounts of money to Bamberg, and started the group by hiring a quite large number of people, and I still had some in Stuttgart (who partially came with me).

I do not regret doing it that way – we had a group that worked well together directly after a couple of months. People help each other, which takes away work from myself. It could have been a bit slower, but that is ok. What I hoped to happen, however, is that a large group size also attracts more (internal and externally funded) projects. So far, I’d say this did not happen and I overestimated this effect. Things are slowly starting to develop, but I do not think that the effect of having a considerably large group contributed to this situation substantially.

Social environment: let it happen vs. actively developing it

I don’t think it is necessary to offer “social events”. Ph.D. students and postdocs are social people and will build something themselves. I do, however, believe that creating opportunities is important. I personally don’t like to have lunch, but we do a group lunch after the weekly group meeting; and my feeling is that it contributes to a good working environment. We further do a monthy bar night, together with another topically related research group. Finally, we do yearly retreats which include social events.

My impression is that it is good that we started these things early. My impression is that we are a healthy research group that works well together. This might have worked out without these initiatives or not. Difficult to say. However, they probably didn’t hurt, so I would do that again.

Group meetings, retreats; embedding in existing environment vs. building something yourself

I found myself in Bamberg in a building that is not the main computer science building, which made interaction across groups challenging. When we arrived, there was no joint meeting across Ph.D. students or a shared colloquium. The latter has been started shortly after we started, also with the contribution by some of our group members. However, I underestimated the importance of a similar research culture and similar research questions for a productive environment. My impression now is that having a “language technology environment” is more important than having opportunities for Ph.D. students in my group to talk to students from an entirely other field in computer science. I’d really like to have a shared environment with linguistics, or cognitive sciences, but starting with a group-internal seminar series to which we invite external speakers was, I think, a good idea. I hope to develop other possible meetups in the future though. This remains to be one of the bigger challenges.

Buy laptops or desktop computers, buy servers, rent compute resources

One challenging decision was also what kind of compute environment to invest in. I decided at the beginning to buy powerful laptops for each group member such that rapid prototyping can be done locally. I also wanted to buy a powerful local GPU server, but the delivery times were terrible, so I bought a comparably small one at the beginning, which I will upgrade soon. Further, we have access to a compute cluster with many GPUs for free.

This setup works well for now. I am not sure if just paying APIs would be an alternative, but definitely the overhead of paying this could be more work.

Other topics?

If you would like to hear my opinions about other experiences that I made, let me know. I would be more than happy to make this blog post cover more topics.

[Download this post as PDF]

Updated: