All News

Meet STELLA: Keynote by Felix Moser on Open-Source Voice AI for Health Research, 1 Oct 2026

About Felix Moser

Felix is a Ph.D. Candidate at the University of St. Gallen under the supervision of Dr. Vinay and Prof. Kowatsch working on GRACE, a voice-based conversational agent that delivers Cognitive Stimulation Therapy to people with early-stage dementia. He holds a Bachelor of Science and a Master of Science in Informatics from the Technical University of Munich, where he focused on engineering software-intensive systems and digital health. He specialized in Machine Learning and AI during an exchange semester at the Technical University of Denmark and was later appointed a Research Fellow at Yale University School of Medicine, where he applied time series analysis to public health decision making and explored the integration of AI into clinical documentation.

About the lecture

This keynote presents STELLA, an open-source research infrastructure for designing, deploying, and studying LLM-based conversational agents in digital health. While large language models have made conversational health interventions technically feasible, their use in research is still limited by opaque decision-making, weak adherence to protocols, and the engineering effort required to run real-time voice interactions with participants. Drawing on the development of STELLA as a modular, self-hosted platform, the talk examines how conversational agents can be made inspectable and controllable. It focuses on a pipeline in which specialized experts run in parallel and are resolved through deterministic, safety-prioritized arbitration. Framed within a methods perspective on transparency and reproducibility, this work emphasizes plan-driven conversation flows, configurable experimental conditions, and full session logging as prerequisites for rigorous evaluation. It also explores how shared, open infrastructure can lower the barrier for researchers to move conversational interventions from prototype to real-world studies.

See also: Moser, F., Vinay, R., Kowatsch, T. (2026) STELLA: An Open-Source Platform for Auditable Voice-First Conversational Health Coaching Agents, In Proceedings of the 8th ACM Conference on Conversational User Interfaces (CUI ’26). Association for Computing Machinery, New York, NY, USA, Article 88, 1–7, 10.1145/3816046.3816293.

We are pleased to invite you to join this guest lecture, which is part of our CDHI Lecture Series Digital Health Forum. Registration is not required. Please be aware that we will be recording this guest lecture and will be making it available in our teaching library. If you have any questions, please contact Sabra Melamed (sabra.melamed@uzh.ch) prior to the start of the guest lecture.

Share this post