"I want to bridge the gap between software and hardware"

Prof. Josipović, what is your main area of research?
I work on electronic design automation: tools and techniques for automated and fast digital circuit design. General-purpose CPUs, like the ones we have on our laptops, are no longer giving us the performance improvements that we need. An alternative are hardware accelerators: digital circuits designed for one particular computing task which they can then perform with high performance and energy efficiency. However, the problem is that designing good-quality circuits is extremely challenging: only a limited number of hardware experts have the required knowledge. Additionally, the design process is extremely slow. So, my group is developing different techniques to ease and speed up this hardware design process, enabling different users to benefit from hardware acceleration.

What brought you to your research field of electronic design automation?
During my Master’s studies in electrical engineering in Zagreb, I applied for an internship at EPFL, and I sort of accidentally ended up in the lab that was working on FPGAs, a particular type of programmable hardware accelerators. And I really enjoyed working with FPGAs! I liked that this field combines different elements of computer science and electrical engineering. I stayed in the same lab for my doctoral studies and continued working in this area.

What fascinates you about electrical engineering?
I always liked physics and math, and I was hesitant between studying physics and electrical engineering. But in the end, I decided on electrical engineering, because I felt it would give me opportunities to explore various topics and open doors to different careers. I enjoyed my entire studies and found both the practical and theoretical aspects of electrical engineering very intriguing.

What is the impact of your research on society? And what are currently the biggest challenges in your field of research?
Since our computational demands are continuously increasing, we have more and more large-scale applications that have to run very fast and under strict energy restrictions. The impact of my research is to enable people in different application domains to benefit from the parallelism and energy efficiency that hardware accelerators offer.

The main challenge in electronic design automation is that the way one thinks about software (e.g., programming in C++) is completely different from how one thinks about hardware design (e.g., programming in low-level hardware languages like VHDL). We need to find a good programming model that is intuitive to users, but also allows them to express important hardware properties and parallelism. Next, our circuits need to consistently provide the correct functionality that the programmer requires; we need to be able to verify the compilation process from software to hardware as well as the circuits that come out. Additionally, the hardware platforms that we have today (e.g., FPGAs) still have some fundamental limitations; we need to rethink our hardware platforms and make sure that they are suitable for very different applications that software programmers require.

Which kind of practical applications is your group working on?
In general, the techniques that we are developing are generally applicable, but we have some interesting use cases that we are currently focusing on. For instance, we are working with physicists from CERN who need to use FPGAs in their collider, but it is challenging for them to exploit these devices in the best possible way and extract the performance that they need. We are helping them by providing compilers and tools that enable them to easily do this job.

What are your long-term goals?
With my research, I would like to bridge this big gap between software and hardware and make hardware acceleration accessible to users of different backgrounds.

How does it feel to have taken up an academic career so young? Could you have considered a different career path?
I enjoyed every aspect of my doctoral studies: research, teaching, and the academic environment in general. I was very sure that I wanted to stay in academia, and therefore, I applied only for academic positions and only selected places that I knew I would like to go to. I'm very grateful that I got this position at ETH, as it was definitely at the top of my list. Of course, it is a big change to go from being a doctoral student to an assistant professor. Mainly, I now have a lot more responsibility than before, as well as new tasks in research, teaching, and mentoring. At the same time, I am doing all the things I liked in my PhD, just amplified, so I like them even more!

Are you collaborating with other people at D-ITET or other departments at ETH Zurich?
Since my research is between computer science and electrical engineering, there are a lot of collaboration opportunities with groups in D-INFK and D-ITET. I recently worked on a project with Gustavo Alonso from D-INFK, I also hope to collaborate with Torsten Hoefler. And from the D-ITET side, I am exploring collaboration opportunities with Luca Benini, and Onur Mutlu, as they are very close to my research field. This fall semester, I was teaching a course with Laurent Vanbever and Roger Wattenhofer, who are also from the TIK institute.

How do you like ETH as a research institution? What distinguishes ETH from EPFL where you did your doctoral studies (in terms of university culture)?
I think ETH and EPFL are quite similar culturally speaking; they are both very international, and the students are driven and smart. All my colleagues at ETH have been super friendly and helpful since I joined. Whenever I have any questions, they are always accessible and find time to chat despite their own crazy schedules. And the students here are excellent, too. Somehow, they even push me because they are so motivated and interested, and it's very pleasant to work with them. I also have the impression that students at ETH are more interested in staying here for their doctoral studies, and I appreciate this continuity very much.

How is your group composed? Are you currently looking for doctoral students?
I currently have two PhD students, and another one will start in September. I'm still hiring and I'm looking for students that can be more from the software side or the hardware side. It would be great to have a group of six to eight students at some point. However, I would like to hire them gradually because it is great to have time to spend with them when they start.

What courses are you teaching?
In the fall semester, I was teaching “Discrete Event Systems” with Roger Wattenhofer and Laurent Vanbever, a fifth-semester Bachelor’s course. And for the spring semester, I designed a new course which is called “Synthesis of Digital Circuits”. It is a Master’s course, that is somehow very related to the research of my group and teaches different techniques to create hardware designs. The first part of the course will introduce the students to the already existing tools and algorithms and techniques. The second part of the course will be more research-oriented about challenges and possible improvements. We hope to interest some students in this field and maybe have them join our group at some point.

Has the Corona pandemic had a lasting effect on your teaching and research activities?
It was a challenging time that pushed us to be more creative in how we communicate and work. One of the things that I think is a positive outcome is that we got more used to these virtual meetings with people who are further away. It is now normal for us to talk every week with people who are on the other side of the planet. Teaching was probably the biggest challenge, as it was difficult to build a connection with the students in a virtual setting. So, I am, of course, really glad that we are back to normal.