What cutting-edge project are you working on right now?
The main research topic in my group is the application of machine learning to guide and synthesize new materials for renewable energies directly in a computer. We are currently involved in the innovation network called ARTEMIS that is focused on machine learning to identify new supramolecular materials for energy and medical applications, and I am one of the coordinators of this project. Supramolecular materials are a broad range of large molecules that can be assembled a bit like Lego bricks to make a variety of geometries with different functionalities. There is a big hope that they can be used for many applications ranging from drug delivery to electrochemistry as the new catalysts. However, identifying the correct ones amid all possible molecular structures is highly non-trivial. Data-driven methods can help in finding patterns across the chemical space for target functionalities.
What was the key experience that made you want to do research in data science and what fascinates you about working with MDSI?
I was attending a conference in Boston when I heard about the importance that data science and in particular machine learning was going to have within the field of material science for the first time. I was fascinated by the use of this class of smart algorithms to teach them chemistry and physics, or at least part of it. It is interesting because machine learning provides a new set of tools that complement the others we already have, i.e. experimental work and numerical simulations, and integrate with them. The MDSI represents a perfect environment to exchange ideas and knowledge about data science although my background is not in it. Thus, having an Institute which summarizes the data science activity all across TUM is extremely important. I am already collaborating with several core members of the MDSI like Prof. Dr. Casini or Prof. Dr. Günnemann.
What paradigm shift do you expect within the next ten years, triggered in particular by the Institute's interdisciplinary approach?
In my field there is no doubt that machine learning is going to change everything deeply. I have heard once that “in the future, machine learning is not going to replace radiologists, but to replace radiologists who do not use machine learning”. I kind of believe that the same applies to my field, in the future all people who are involved in material science and device fabrication will use some machine learning methods. This poses an important challenge: to educate students and researchers in machine learning, considering that they mostly have a background in physics and chemistry or engineering in my field. In this perspective, an institute like the MDSI can pave the way to coordinate this crucial activity.
And what else? (Wishes, curious moments in research, outlooks, future projects etc.)
Regarding future projects, they are associated with building on top of the current ones in order to develop better algorithms for materials. Most of the algorithms we use have been developed within fields that are not related to material science (like image recognition or text analysis) and I wonder what could be the perfect architecture of an algorithm for understanding material properties.