Professor Zhang, welcome to ETH

We welcome you at ETH, Prof. Zhang. What led you to decide to begin a professorship at our university? Please describe your initial impression of ETH when you came here for your interview.
Thank you! I am looking forward to starting the next phase of my academic career here at ETH. My first impression of ETH, to be honest, was how beautiful it is. The day before my interview was a very sunny day, and when I stood on the terrace of the main building that overlooks the lake and the old town, the view was just breathtaking. During my interview, I found all the characteristics that I was expecting from a top university and a top computer science department. There are leading researchers in almost all the subfields of computer sciences. I was excited about the collaborations I could conduct here at ETH. My student meeting also went really well, and I was impressed by the students’ competence and enthusiasm for what they are doing. Another impression is the top-down push for data sciences, such as the data science initiative in Switzerland. These efforts will definitely attract more students with research interests that match with what I hope to achieve. All of this makes me very excited about the opportunity to join ETH as a junior faculty member. So here I am!

How do you intend to bring in your background and experiences to strengthen our department?
I hope to be the bridge connecting three existing strengths of ETH: the system group, the machine learning group, and scientists in other departments such as biology, pathology, and social sciences. Computer science is a relatively young field compared with many other physical and social sciences. However, it never stops reshaping the way that scientists in other domains look at their own problems through the “computational lens.” Recently, there have been advances in a range of fields, which I believe will enable the next exciting wave of such reshaping. Machine learning techniques have been improved significantly, and we see in many applications an automatic algorithm become competitive to human quality. Hardware has become much faster, greener, and cheaper, and human-computer interaction techniques have matured significantly. We can start thinking about how to build not only a data system but also data science ecosystems, frameworks, and interaction models to help other physical and social sciences. These types of interaction have existed at ETH for a long time already, and I hope I can further strengthen them.

What are your current research interests?
Right now, one important problem in data science is that many of the techniques needed to unleash the next big thing are available but still far from accessible. For example, just last month, we published a paper in Nature Communications that described how we can build an automatic algorithm for lung cancer diagnoses that can often match the accuracy of experienced pathologists. The technique we employed is quite standard from a machine learning point of view—classifier, feature engineering, etc. But it still took our collaborators, who are pathologists and medical doctors, a considerable amount of time to put these together in the current ecosystem of data processing. In this case, the enabling technique is available but not accessible enough to our pathology collaborators. My current research interest is to understand—and try to close—this gap between the availability and accessibility of machine learning techniques. Of course, there will always be a gap, but can we make our current machine learning ecosystem much easier to use? I hope my research can answer this question. I expect that my research group will continue working with experts in diverse domains to build data-driven applications, extract common patterns from these applications, and build generic platforms and systems to support them that are faster, greener, more scalable, and more user-friendly.

How do you feel your teaching style can benefit our student population?
For classroom teaching, I believe that it is important for students to not only understand the technique itself but also how such a technique already helps or could help a real application. In my opinion, understanding the interaction between the application and the technique is very important, and ETH has a long tradition of emphasizing this interaction. I hope my experience in working with a range of domain scientists could provide some new examples for students.

What do you look for in a doctoral student who wants to work under your supervision?
Doing research is fun! My greatest hope is that my students share the enthusiasm and passion for research. I would hope they keep a clear career goal in mind, are willing to work as hard as they can, and have fun while making it happen! I was very lucky as a PhD student that my advisor put me in a position where I could work closely with a range of domain scientists, such as palaeontologists, geologists, biochemists, pathologists, and so on. This experience shaped the type of research that I want to do. As an advisor, I will try my best to create the same kind of environment and opportunity for my students. In return, I hope my students are willing to go the extra mile in understanding these applications and domains, feel the pain and frustration that our users are experiencing, and think hard about how we, as computer scientists, can help. I am confident that all ETH students have the technical competence to succeed. So the important thing is to make sure we have a matching research interest and shared vision for data systems as a major driving force of our time in making the world a better place. On my homepage, I have put a list of more than ten example topics that I am excited about to help potential students know my interests better.