Couldn’t make it to SEMANTiCS 2025 in Vienna? Our Community Growth Manager, Larry Swanson, was there soaking up the talks and hallway conversations with fellow semantic-tech enthusiasts. From real-world applications like ZEISS’s Service Copilot and IKEA’s evolving enterprise-scale knowledge infrastructure to recurring themes around the interplay of knowledge graphs and LLMs, Larry came back with plenty of inspiration and a few standout takeaways. Keep reading for the full report!
metaphacts' SEMANTiCS 2025 Conference report by Larry Swanson
I had an amazing and edifying time at the SEMANTiCS 2025 conference in Vienna recently.
As always, the "corridor track" was my favorite place to learn and connect. I talked with dozens of fellow semantic-tech nerds about everything from knowledge graph quality assurance to stakeholder evangelism, from network science to autonomous vehicles, from biological ontologies to industrial workflows. It's exciting to see semantic technology being applied to such a wide variety of use cases, and it's great to see so many brilliant minds tackling the important work of semantics.
I spent an entire day after I got home compiling and reviewing notes and following up on LinkedIn and email connections, and I'm nowhere close to being done. Too many great people...
The conference program was just as rich and engaging as the hallway track. I typically cherry-pick three or four sessions a day at a conference like this. Applying the same heuristics as usual, I ended up in seven or eight sessions each day. Too many great talks...
A recurring theme in nearly every talk (and in many of my hallway conversations) was the interplay between knowledge graphs and LLMs. The insights that AI engineers have gleaned over the past few years are now being applied in more fully intelligent AI architectures - KGs providing grounded facts and logical reasoning and LLMs building and maintaining ontologies, writing queries, and mapping database schemas to ontologies. "Synergy" is an overused word, but I think it has earned its place in hybrid-AI conversations.
Every talk had at least one inspirational or actionable take-home, but a few talks stood out.
As a long-time omnichannel service designer, I was extremely interested in Max Gärber and Sonam Chugh's talk on the ZEISS Group's Service Copilot. It's an AI-powered assistant that gives field engineers real-time, actionable insights informed by an industrial knowledge graph and navigated with conversational chat agents. The service is built on an agent-orchestrated modular architecture that integrates content, service, planning, and recommendation capabilities. It's currently live and in production, serving about 1,000 users. Semantic engineering lies at the core of the system, with a knowledge graph connecting concepts and instance data and the ontology serving as the point of integration with LLMs. It was great to see a real-world example of knowledge-driven agentic AI.
IKEA's lead ontologist Christelle Maignan talked about the evolution of their knowledge graph capabilities from KG-driven product recommendations to an enterprise-scale decentralized knowledge infrastructure. The goal for the next iteration of their KG is to give customers the detailed product information they need to make informed, sustainable choices. This is a big step forward from their earlier SME-informed recommendation system and will require collaboration across a number of new business domains. There were many great details and examples in this talk about the rationale for decentralization, the importance of aligning domain experts on a core ontology, how to govern metadata about semantic assets, and the importance of an internal communication strategy. As a practice democratizer, I really appreciate how transparent the IKEA team is about how they're evolving their KG practice.
My favorite session take-home, though, was Founding Director of the Amazon-Illinois Center on AI for Interactive Conversational Experiences (AICE), Heng Ji's answer in her keynote on Thursday morning to the question prompted by the ubiquity of LLMs, "Are we [knowledge engineers] dinosaurs?" Her answer, which was convincingly supported in her talk, a resounding "No!" Long live semantic technology!
The conference was extremely well organized and run. Kudos to Sahar Vahdati, Amin Anjomshoaa, and the rest of the SEMANTiCS organizing team. A similarly capable team is already planning next year's event. Looking forward to seeing everyone in Ghent, Belgium, in 2026.
Your knowledge-driven AI platform
Ready to leverage AI for your enterprise? metis is a knowledge-driven AI platform combining large language models & knowledge graphs to deliver AI agents that provide generative power, semantic precision & contextual, explainable insights for your business.
At its core, metis is grounded in a sophisticated semantic model that captures essential context and expert knowledge from domain specialists and business users. This unique foundation facilitates a powerful human and AI collaboration, including an augmented intelligence and the human-in-the-loop approach, which leverages human expertise and experience at the heart of all AI interactions.
See metis in action and request a demo today!
