Graphical User Interface Agents and LLMs for Recommender Systems: My 2025 Conference Talks
I had the opportunity to speak at two conferences in 2025: AI Engineer Europe in Paris and Data Innovation Summit in Stockholm. The talks cover two different sides of state-of-the-art applied AI: building autonomous agents that interact with graphical interfaces, and the impact of large language models on recommender systems.
AI Engineer Europe 2025, Paris: Graphical User Interface Agents
AI Engineer is a high-signal technical AI conference where engineers share frontier knowledge and showcase world-class work. The first European edition was held at STATION F in Paris.
In this talk I explore how GUI agents (including computer use agents, as popularized by Anthropic) enable AI systems to interact with real-world user interfaces like a human would. I cover why GUI agents matter, the trade-offs between API-based and GUI-based approaches, how hybrid systems combine the best of both, and share insights from building a voice-controlled in-car GUI agent at BMW Group Research.
Talk Recording
LLM-Based GUI Agents: Bridging Human Interfaces and AI
Slides
Resources
- API Agents vs. GUI Agents: Divergence and Convergence — Zhang et al., 2025 (Microsoft)
- Mobile-Agent-v3: Fundamental Agents for GUI Automation — Ye et al., 2025 (Alibaba)
- Small Language Models are the Future of Agentic AI — Belcak et al., 2025 (Nvidia)
Data Innovation Summit 2025, Stockholm: Power of LLMs for Recommender Systems
Data Innovation Summit is the biggest Data, AI and Advanced Analytics event in the Nordics.
In this talk I explore the use of large language models for recommender systems, focusing on how advancements in LLMs have transformed traditional recommendation paradigms. I cover the benefits and challenges of integrating LLMs into recommendation frameworks, along with practical research insights and LLM-inspired innovations from industry leaders in this space like YouTube, Spotify and Netflix.
Talk Recording
Power of LLMs for Recommender Systems
Slides
Resources
- Matrix Factorization Techniques for Recommender Systems — Koren et al., 2009
- Deep Neural Networks for YouTube Recommendations — Covington et al., 2016 (YouTube)
- Recommendation as Language Processing (P5) — Geng et al., 2022
- Recommender Systems with Generative Retrieval — Rajput et al., 2023 (Google)
- Better Generalization with Semantic IDs — Singh et al., 2024 (Google)
- Improving Content Retrievability in Search with Controllable Query Generation — Penha et al., 2023 (Spotify)
- Encouraging Exploration in Spotify Search through Query Recommendations — Lindstrom et al., 2024 (Spotify)
- Foundation Model for Personalized Recommendation — Hsiao et al., 2025 (Netflix)
- Sliding Window Training for Foundation Models — Joshi et al., 2024 (Netflix)