“Attending DAC was like winning the lottery.” An interview with Ondřej Galeta about the major conference, his PhD and the scientific community

Ondřej Galeta is a new PhD student at the Department of Computer Systems, Faculty of Information Technology, Brno University of Technology; his supervisor is Prof. Lukáš Sekanina. In his research, Galeta seeks to combine machine learning and evolutionary algorithms as two approaches that can be used to design an approximate circuit. Although he has only just completed his master’s degree, he has already had the opportunity to meet the leading researchers in his field and even present his own work to them. In April, he was accepted into the Young Fellows programme, which forms part of the DAC conference (The Chips to Systems Conference; formerly: Design Automation Conference). DAC is one of the world’s leading conferences focused on the design of electronic circuits and systems and their automation (EDA) – the acceptance rate for DAC papers is around 20 per cent. The conference bridges the gap between academia and industry and showcases the latest research, technologies and tools in the field of chip design, hardware and complex electronic systems. This year’s event took place at the end of July in Long Beach, California.

Several DAC conference activities are regularly aimed at students – the main one being the Young Fellows programme. The programme is intended for students nearing the end of their master’s degree and at the start of their doctoral studies. It consists of selected conference content, the aim of which is to bring talented students at the start of their academic careers directly into the field’s exclusive scientific community. For someone on the cusp of doctoral study, such an experience and opportunity is unique. That is precisely why we invited our successful student for a short interview.

***

How would you describe the research topic you’re interested in to a layperson?

I’ve always been fascinated by artificial intelligence, but I must admit that when I started at FIT, I viewed neural networks and classifiers as a kind of magic. I didn’t understand them. At the same time, I’ve always been interested in hardware design. The automation of electronic circuit and system design, which is my area of specialisation, brings together these two seemingly disparate fields: Using artificial intelligence algorithms, we aim to design approximate circuits – that is, circuits that represent logical functions in, shall we say, a more relaxed form. Such circuits operate with a certain acceptable degree of approximation, which makes it possible to speed up and streamline the system’s operation. Put simply, I am trying to combine machine learning and evolutionary algorithms as two approaches that can be used to design an approximate circuit. During my bachelor’s degree, I learnt about the work and achievements of the EvoAI Hardware research group, which led me to this topic. My stays in Stuttgart and Crete as part of the Erasmus programme then helped me to specify how to develop it further and take it to the next level.

It could be said that you’re working on a topic that isn’t exactly being pursued by hordes of scientists around the world. Is that correct?

In a way, you’re right. When writing my dissertation, I didn’t have many texts to draw on. There are only a few articles available. The approach I’m primarily focusing on isn’t, of course, the only one possible – currently, the use of large language models for design is popular. My approach is more rudimentary, more granular. In fact, I’m creating my own neural network and using it to try to speed up the evolutionary algorithm in design.

In what way is your approach more advantageous or more interesting for science?

We want to combine the best of both approaches. Evolutionary algorithms offer an excellent methodology for rapidly generating a large number of circuits, then evaluating and selecting them. And machine learning can be integrated into the part of the evolutionary algorithm where it proposes circuit modifications based on experience from previous solutions. The evolutionary algorithm then utilises these suggestions when creating further circuits. Our approach is relatively new; we rely heavily on modern insights into neural networks.

You must have been delighted to be invited to the DAC conference programme…

Absolutely. It’s a prestigious event, and not just because of the CORE acceptance rate. The conference is unique in that it has a dual programme – one for the academic community and one for industry. Top engineers engage with academics. And on top of that, DAC offers programmes to support early-career researchers. The organisers aren’t afraid to invest in people trying to break into the field. And this year’s bonus: it all took place in a beautiful location, Long Beach, California.

How on earth did you manage to get into such an event?

The idea came about through discussions with Professor Sekanina. Of course, you need a letter of recommendation, and support from the faculty is key in that regard – and Professor Sekanina’s signature, given that he also has a decisive say at the sister conference DATE, was crucial for my participation. Given how many people applied and how many were ultimately selected, I regard my participation as winning the lottery. And as a great honour. I didn’t personally meet any other Czechs there, which of course doesn’t mean we don’t have capable engineers back home; it’s more a matter of chance. And what’s interesting is that whenever I mentioned Brno, it was often followed by a question about Vojtěch Mrázek. It seems that Associate Professor Mrázek is well-known abroad.

What do you consider to be the highlight of your participation in DAC?

The conference lasted several days. As part of the Young Fellows programme, we had our own schedule, which focused, for example, on strategies for publishing articles and on meeting senior researchers and scientists. We also had tasks to complete; for instance, we prepared short videos on the topics of articles that had caught our interest. A great deal of the focus was on networking. You meet people from prestigious global companies and even Nobel laureates – in my case, the American physicist John M. Martinis. These meetings are definitely a bonus – I know from personal experience that trying to get in touch with big names, for example by sending them emails, is a difficult process. Here, you meet them in person and they’re ready to talk to you. Finally, I mustn’t forget the opportunity to present my own research via a poster – mine was based on my Master’s thesis. It’s actually incredible that you get the chance to present your ideas at such a major event. People from Apple and Nvidia stopped by my poster… I really appreciate it. Young Fellows is a way for early-career researchers to gain easier access to an environment that is otherwise becoming increasingly competitive and less and less forgiving. If I’m lucky enough to present a paper at such a major conference in the future, this year’s participation will certainly help me.

Paradoxically, did the event leave you with the impression that the world of top-tier scientists is so exclusive that, unless you’re lucky enough to get into an event like DAC, it remains off-limits to you?

From my current perspective, it seems to me that a PhD is the key. If you want to work in top positions at major companies or set up your own business with genuine international ambitions, a PhD is one of the easier routes. By the time you’ve completed it, you’ll already have several publications to your name and a track record to show for it. And, above all, it’s clear that you’re capable of learning and devising innovative approaches. In my view, it’s a better springboard than starting out after a master’s degree by gaining work experience in less relevant companies and opting for repetitive work, which is, moreover, currently under threat from artificial intelligence.

Did the DAC influence the direction of your research or your vision for your professional career in any way? Have you taken any lessons away from it?

I wanted to see how the top scientific and industrial communities operate – how those people think and how they communicate. The conference covered a wide range of topics, from highly specialised transistor design right through to design using machine learning and large language models (LLMs). And the trend towards AI agents dominated everything. More than half of the papers presented focused on some form of artificial intelligence and machine learning. Incidentally, a great deal of effort in the field today is directed towards making the operation of neural networks more cost-effective. It’s therefore natural that very few people understand the full technical scope of such an event – and it was interesting to observe how people coped with a certain degree of unfamiliarity. The conference was also helpful in terms of content; I identified two or three further questions and obtained contact details for people with whom I can discuss them further. I saw new, different approaches that I had previously thought might not even be possible. It was a great experience and I’m glad I went.