Keynote Speakers
Santeri Auvinen is is a Manager in Software Development at Elisa, with over 25 years of experience leading product development, engineering teams, and complex R&D initiatives in the technology sector. A hands-on practitioner since the late 1990s, his background includes a significant tenure at Nokia and Microsoft, where he led software design for multiple large-scale product lines and managed global technology transfers. Most recently, he has focused on leading engineering teams for major consumer television and streaming services at Elisa.
Santeri is deeply passionate about bridging the gap between industry execution and academic research. He is a core member of his organization’s university collaboration team and actively works with students to explore emerging technologies. His engineering focus centers on sustainable digital product development and green coding, driving this domain forward through close collaboration with university ecosystems as an industrial supervisor for student research and Master’s thesis projects.
Keynote #1
Grounding AI: From Probabilistic Models to Mission-Critical Software
Abstract
While foundational models are advancing rapidly, industrial deployment reveals a critical friction: a model by itself is simply infrastructure that solves no business-level problems out of the box. Production success requires identifying specific bottlenecks where technology adds measurable value, mastering integration with legacy environments, and maintaining realistic software engineering cycles.
This keynote provides a practical, industry-focused overview of how software organisations deploy these technologies, structuring the landscape into three application categories: AI as the Application (user-facing custom agents), AI for the Application (background engine-room orchestration), and AI for Making the Application (developer velocity and automated testing tools). Drawing from firsthand experience as a Software Development Manager and industry thesis supervisor, Santeri will examine the operational risks of each category and explain why unmanaged open-source models, without deep internal integration capabilities, create long-term maintenance liabilities.
The keynote also addresses a critical architectural challenge: how to decouple traditional deterministic software and stable user experiences from the non-deterministic, probabilistic nature of AI models.
Finally, the talk outlines a vision for restructuring future product development teams. Drawing on observations from university thesis projects, Santeri argues that successful technology transfer depends on a new role: the “Deep Scientist”, where team members responsible for building the validation frameworks and guardrails needed to anchor fast-changing, non-deterministic models into stable, mission-critical production systems.
Filip Ginter is Professor of Computer Science at the University of Turku, Finland, co-founder of the TurkuNLP research group, and a faculty and management member at the ELLIS Institute Finland. His work focuses on core NLP methodology, data resources and their evaluation, high-performance computing, and large-scale multilingual and cross-lingual NLP, with particular interest in digital history research. He has contributed to many foundational NLP models and tools for Finnish, as well as to international efforts such as Universal Dependencies. Through the ELLIS Institute Finland and the national AI-DOC doctoral program in AI, he also helps advance Finland’s national AI ecosystem in general.
Keynote #2
AI for Digital History: The Algorithmic Challenges and Opportunities in Tracing Meaning at Scale
Abstract
Recent progress in AI and natural language processing (NLP) is reshaping what is possible in research. While public attention often focuses on chat-based AI, some of the most transformative opportunities lie in the high-performance computing–driven use of modern models on large and complex datasets. This is especially true in digital humanities and digital history, where computational methods can now address research questions that were previously beyond reach. Digital history is both scientifically important and computationally fascinating. It brings research questions that are rarely straightforward, alongside domain-specific challenges such as historical language variation, digitization noise, and massive multilingual text collections spanning centuries. Among the most compelling new directions is the emerging ability to trace meaning, i.e. how ideas spread, evolve, and influence one another across languages, regions, and historical periods. Modern AI and NLP models are beginning to make such analyses possible at scale, opening new ways of studying cultural and intellectual change. In this keynote, I will discuss key advances and the many remaining challenges in applying AI and NLP to digital history, with particular emphasis on large-scale cross-lingual tracing of meaning. I will argue that this domain is not only a valuable application area in its own right, but also a source of methods and perspectives that may prove inspiring beyond the digital humanities.