Background: Brings deep technical expertise from particle physics research where he applied AI to real-time applications processing enormous amounts of data at unprecedented speeds. This background in high-performance computing and real-time inference—conducting millions of operations per second on specialized equipment like FPGAs and DSPs—provides unique insights into the infrastructure challenges facing enterprise AI deployment. At Swisscom, he leads the Digital Lab at EPFL, fostering long-standing collaborations between industry and academia on AI topics including explainability and transformer architectures.

Dobos has witnessed the evolution of AI from its early applications in particle physics to its current enterprise implementations, giving him a rare perspective on both the technical and business transformation potential of AI systems. His work bridges the gap between cutting-edge research and practical industrial applications, particularly in telecommunications infrastructure where reliability and performance at scale are paramount.

Key Perspective: Dobos defines open source through the rigorous lens of scientific reproducibility, emphasizing that true openness requires complete transparency including data filtering processes and methodological documentation. « From an academic point of view, open means reproducible. And everything what is needed to reproduce results would basically fall into my definition of open, which includes even filtering of the data. » This scientific rigor reflects his particle physics background where experimental reproducibility is fundamental to valid research.

His perspective on AI infrastructure development draws direct parallels to internet evolution, providing historical context for current debates about open versus closed AI systems. « We have seen the development of Internet where basically open source was winning over long term. I think nobody would seriously try to publish a non open source browser today or some kind of Internet infrastructure which would be not open source. » This historical analogy suggests that AI infrastructure will follow similar patterns toward openness and standardization.

Industry Leadership: Through his role leading Swiss AI weeks, Dobos has demonstrated remarkable ability to build consensus across diverse stakeholder groups. The initiative successfully brought together 55 organizations spanning government, private sector, and academia to test Swiss open source models. « We collected 55 organizations, we have been talking to at least 55 CEOs plus a little bit more and try to explain them exactly why does it need AI sovereignty? And surprisingly we didn't have to explain this too much. »

This success reflects his skill in articulating complex technical concepts to business leaders while highlighting practical benefits. The rapid adoption demonstrates that Swiss organizations understand the strategic value of AI sovereignty without requiring extensive technical explanations. Dobos emphasizes that organizational leaders quickly grasped how sovereign AI solutions provide independence, build trust, and offer strategic flexibility in an uncertain geopolitical environment.

Talent Ecosystem Analysis: Dobos provides compelling analysis of Switzerland's competitive advantages in global AI development, identifying talent density as the crucial differentiating factor. « Almost every big AI company that you think about will now basically open a research lab in Zurich or here in Lausanne. There are three reasons for that: talent, talent, and talent. The density of AI talent in Switzerland is amazing. »

His perspective, informed by recent visits to emerging AI hubs including Singapore, China, and Dubai, positions Switzerland's talent advantages within global context. While other regions offer different strengths—Dubai provides vision and unlimited funding, various locations offer favorable regulatory environments—few match Switzerland's concentration of technical expertise built through decades of excellence at ETH Zurich and EPFL.

Dobos acknowledges areas for improvement in the Swiss ecosystem, particularly around risk-taking culture and funding accessibility. « We are not very good in taking risks. I guess as an entrepreneur you will have also made the experience with maybe getting funding in getting an investor in Switzerland. It's not as easy as other parts of the world. » However, he argues that Switzerland's talent foundation provides sustainable competitive advantages that capital alone cannot replicate.

Technical Infrastructure Vision: Drawing from his telecommunications and supercomputing background, Dobos emphasizes the critical importance of infrastructure reliability and performance in AI deployment. His experience with Swisscom's massive data processing requirements provides practical insights into scaling AI systems for enterprise applications. He understands that successful AI implementation requires not just advanced algorithms but robust, reliable infrastructure capable of handling production workloads.

His work with EPFL demonstrates how industry-academia partnerships can accelerate AI development while maintaining research excellence. The collaboration has produced practical advances in explainable AI and transformer architectures while training the next generation of AI talent. This model of sustained partnership between industry needs and academic research could serve as a template for other regions seeking to build AI capabilities.

Regulatory and Trust Framework: Dobos articulates how open source AI development aligns with European regulatory requirements and business culture. The emphasis on transparency and reproducibility in open source models provides natural compliance advantages as AI regulation evolves. For Swiss organizations, where trust and reliability are premium values, open source approaches offer transparency that closed systems cannot match.

He positions AI sovereignty not as technological nationalism but as practical business strategy. Organizations need reliable access to AI capabilities regardless of geopolitical tensions or commercial disputes. « Having that choice is worth having such a model. This is why up to now 55 organizations joined and said yes, we want to start building on this and each organization will make its own decision if it's their choice A, B or C in their planning. »

Future Applications: Dobos envisions AI applications extending far beyond current language model use cases, particularly in critical infrastructure and specialized technical domains. His background in particle physics and telecommunications provides insights into how AI can enhance complex technical systems while maintaining reliability and performance standards. He emphasizes applications in synthetic data generation for testing edge cases and improving system stability under unforeseen circumstances.