“What we’ve learned along the way”: The Computational Carbon Chemistry Group says good-bye…

Jim Gray is one of the smaller meeting rooms at HITS, where all seminar rooms are named after renowned researchers and pioneers in the computer sciences, and it is with mixed feelings that I am here today to interview three group members of the Computational Carbon Chemistry group. After an amazing five years as a Junior Group at HITS, they are leaving according to schedule for pastures new and in case of group leader Anya Gryn’ova also foreign at the end of May. Before their departure I have a chat with Anya, her PostDoc Christopher Ehlert and PhD Student Anna Piras to hear stories from throughout their time at HITS and what they have seen and learned along the way…

Anya: My name is Anya Gryn’ova. I’m a Junior group leader of the Computational Carbon Chemistry group. I’m originally from Ukraine. I did my PhD in Australia and my postdoc in Switzerland. Five years ago, I moved to Heidelberg to start my independent research career. This stage of my life is now coming to an end, because in April, me and some of my group members will be moving to the University of Birmingham in the UK to continue our research there.

Christopher: My name is Christopher Ehlert, I’m a PostDoc in the Computational Carbon Chemistry group. I joined in October 2019. After four and a half years, my stay here at HITS will also come to an end and I will move to industry.

Anna: My name is Anna Piras, I joined the group in October 2019 as a PhD Student. My PhD will be completed in a couple of weeks and I will go back to Italy for a PostDoc position.

Often scientists are asked about special moments in their career. What do you consider your scientific highlights during your time at HITS?

Anya: There were many such exciting moments when I discovered or understood something, something that I did not expect. But there are two things that stand out for me, relating to something I started appreciating over time. One is how much I have learned from my group members and how inspiring they are. I’ve often thought “how did they come up with this idea?” or “how did they manage to do this?”. This is so amazing! The other aspect that also unveiled for me over time is when I joined HITS, I could never imagine myself being anywhere near method development. And even though the group is still mostly oriented towards applications, we do have several projects that are targeting more methodological aspects of modeling. I think for this we must give credit to HITS for encouraging me to pursue these research avenues.

Anna: When I started my doctoral research, my project was supposed to be something very applied, and we wanted to use methods already available in a novel way to answer some new questions. The big moment for me here at HITS was when we discovered that this didn’t work, and we had to fully redesign what was my PhD. So, there was this big change, and it was a bit scary to find that some of the work I had done was not valid and we had to do something new, but this was made possible by the freedom we had at HITS. It was a significant moment, and it allowed me to understand that science is not linear and sometimes you have to accept that something you did is just wrong.

Christopher: For me the fact stands out that we have been dealing with such a broad range of topics. One highlight was the electrochemical part that we have been working on, which also has a relatively high relevance in real applications. We also had many fruitful projects with experimentalists and a lot of topics I would have never thought I would work on, for example all the machine learning applications that I became interested in.

Would you say that due to these things you have also grown as a person?

Anya: I would like to think that I have. For me these past years were a transition from a postdoctoral researcher who did mostly scientific research work to a group leader, which involves managing people in different roles: as a mentor, as a supervisor, and most of the time as a colleague and collaborator, building personal relationships. That is something that was not easy or natural for me, but I think I learned how to create a group that is united not just by the same workplace or the same line of reporting, but by some personal relationships as well.

Anna: For me it was quite hard at first, especially because I come from Italy and the environment there at the university was still very conservative. So, when I arrived here, I really struggled to admit failure. Being wrong was a big problem. It took me a lot of work and discussions with Anya. For a long time, I tried to solve the problem on my own and I didn’t ask for help, because that would have meant to admit failure. It took me a lot of time and work to accept that some of my work was wrong and I had to redo it. I don’t think there was really a strategy, but it was a matter of being strong and brave enough to admit it. That was for sure a big personal growth for me here at HITS.

Christopher, you mentioned that the unexpected played a role in your career here at HITS, that you suddenly had to take on things you hadn’t calculated before, machine learning for example.

Christopher: Indeed. Before I came here, machine learning was something I was not interested in at all. It was like a black box or even dark magic (laughs). I thought there was no scientific background behind it. But then we had PhD students who were really into it, and they somehow convinced me that there is more behind it. Now, I’m developing neural networks and machine learning models myself. I’ve completely changed my view on this topic. When you do a quantum chemical calculation, for example, you usually find a solution iteratively. You need to solve an equation again and again, until it’s what we call self-consistent. The result you obtain is, for example, the electron density of a molecule. What can also work with machine learning is that you enter a description of a molecule into a machine learning model, and it immediately spits out the density, so you avoid all these iterative processes. This of course saves a lot of computational time and makes it way more efficient.

Anya: I had absolutely the same opinion as Christopher. At the very end of my PostDoc, the group I was working in started to use machine learning and I thought about how we overrate something, which is essentially just fancy regression. The scientific side would be lost by just processing huge amounts of data without actually understanding anything. I was so certain that I would never want to use machine learning. I thought it was just a hype that will find its applications but would not lead to any deep understanding. I had exactly the same trajectory as Christopher: Understanding better what machine learning is, understanding its limitations. In the end it’s anything but a magical black box and I see the value that it can bring to actual knowledge, fundamental physics and chemistry. What really blew my mind in terms of machine learning were the many capabilities of both kernel based methods and deep neural architectures. Sometimes you can really see the patterns in the properties just from the behavior of the model. I’m very excited about one of our projects. It’s a technique, I would even say a bit of a philosophy of looking at chemical space, millions and trillions of possible molecules, and trying to both organize this chemical space and explore it with a specific practical question in mind. We came up with a very elegant and conceptually simple way to do both, which allows us to predict properties of molecules from very condensed information. Typically, we would represent a given molecule to a machine learning model with thousands of numbers and get reasonably accurate predictions but now we can represent every molecule with two or three numbers and get better predictions. I think this is mind blowing. What makes it even more exciting is that we transferred this concept to fields completely outside of chemistry, image and voice recognition for example. I think this is a very general concept and I’m extremely excited about it.

We talked about interdisciplinarity here at HITS. You have also worked with Jan Stühmer, group leader of the junior group Machine Learning and Artificial Intelligence which was founded at HITS in September 2022.

Anya: Yes, this is a project which one of my PhD students is working on. He is using various graph neural network architectures to predict redox properties of molecules and hybrid materials. This is where I’m completely out of my depth in terms of technical expertise and having a machine learning expert at HITS is great. We get a lot of technical advice from Jan, and I can only say that those years that passed between the position being announced, and Jan joining were spent in anxious anticipation of when he finally joined, and we were able to pick his brain (laughs).

I’m sure he’ll be pleased to hear that.

You all came to HITS from different countries, with different backgrounds. How would you describe HITS to somebody who doesn’t know it?

Anya: I remember when I visited HITS for the first time for my interview and talked to the management, other group leaders, and members of the institute, some of them told me it’s all based on Klaus Tschira’s vision for this open flow of ideas, and “thinking beyond the limits”. And I thought to myself these are all nice catchy phrases that I guess people are supposed to tell you when they are recruiting you. But I came to appreciate that they really hold meaning. I think HITS is special in many ways, also due to its setting and location which is not typical for a research institute. Normally, it doesn’t really matter where an institute is located. But the way HITS is set up, I really feel like the mind can wander across the scientific landscape. There is this freedom and courage to explore things.

Anna: For me it’s also the incredible interdisciplinarity, despite the fact that we are such a small number of people (editor’s note: approx. 150). There are so many topics that get explored, it’s really something unique. I think it helps also because when you meet people you don’t always talk about work, but at the same time you understand more about your work in context because you talk to people who come from different backgrounds. And it’s also that it’s very international, so you have a lot of diversity on these aspects.

Christopher: I would also add that the location is very special. The scenery, the garden, the park. Compared to a place like the university it’s also much calmer, there are no students running around, unless you invite them to work with you (laughs). All that gives you a work environment where you can focus on your work, and you’re not interrupted throughout the day by stuff that it not related to your research.

My final question is about your personal outlook for the future. In broader terms, what are you most looking forward to?

Anna: I really liked to be in Germany and at HITS, but I’m also looking forward to going back to Italy. For me, it’s important to work in my home country and make my contribution there. I’m looking forward to advancing another step in my academic career and see how I will find my feet in an environment so different from HITS.

Christopher: I will go for a job in the industry which will have nothing to do with chemistry. I will work on a lot of new topics and that’s what I’m looking forward to. It will also have to do with a lot of data, data analysis and data sets but with a different topic. The way of addressing the problems, the analytic techniques, may probably be the same, though.

Anya: I will be among other chemists again. The interdisciplinary nature of HITS is a great thing, but we were the only chemistry group, so I’m very much looking forward to being part of a school of chemistry which now has about 60 research groups across all fields.

Is there anything you will miss most about HITS and Heidelberg?

Anya: Anna and Christopher and the rest of my group (laughs). Only a few people are coming with me to Birmingham, all others are continuing elsewhere. So I will most miss my group.

Anna: It’s the same for me, I will really miss my colleagues. We had a really good time together, both at work and personally. I met so many great people at HITS, I have great friends, people I consider family. I will miss that for sure.

Christopher: Yes, I’ll also miss my group members and colleagues. Besides that, I will miss the scenery that changes with the seasons. Heidelberg is always beautiful: in summer, winter or autumn. And I will miss this place here of course. The good food in the canteen, and the great coffee.

Thank you all very much. And all the best for you as you are pursuing your own career path separately. I hope you will always feel connected as a group.

More about the „Computational Carbon Chemistry“ group: https://www.h-its.org/research/ccc/

More about the „Machine Learning and Artificial Intelligence“ group:

https://www.h-its.org/de/forschung/mli/.

HITS