ODD-First Agent AI for Scalable, Traceable Autonomy Scenario Libraries
30 April 2026 | 2 p.m. CEST | online
Deontic is an agentic AI platform that transforms ODD definitions and engineering requirements into simulation‑ready scenario libraries with full traceability and validation integrity. Powered by a fine‑tuned LLM, Deontic automates the authoring and parameterization of scenarios across road geometries, traffic controls, actor behaviors, weather conditions, work zones, and complex long‑tail interactions. As standards evolve, it continuously updates coverage and documentation - keeping development teams audit‑ready at all times.
AGENDA
30 April 2026 | online
2:00 PM CEST
Welcome & Introduction
Alexander F. Walser | ASCS e.V.
2:05 PM
ODD-First Agent AI for Scalable, Traceable
Autonomy Scenario Libraries
Yves PEIRSMAN | Deontic
2:45 PM
Q&A
Ask the speaker
3:00 PM
End
SPEAKER
Yves PEIRSMAN
Yves Peirsman is a Natural Language Processing (NLP) expert with over 20 years of experience across academia, industry, and entrepreneurship. He conducted research at KU Leuven and Stanford University before moving into industry roles at Textkernel and Wolters Kluwer, where he developed advanced NLP solutions across multiple sectors.
As founder of NLP Town, he advised organizations on applying NLP to tasks such as search and classification. He is also an active open-source contributor, with widely used models downloaded millions of times, and organizes the Belgium NLP Meetup.
Currently, Yves is CTO at Deontic, where he applies AI to improve compliance processes. He is known for combining academic depth with practical impact in real-world NLP applications.
TARGET AUDIENCE
- ADAS and autonomous vehicle engineers
- Requirements and systems engineers
- Simulation and scenario engineers
- R&D and innovation leads
- AI and machine learning practitioners in automotive
- Functional safety and validation experts
- Researchers
KEY TAKEAWAYS
- ODD-first approach: Scenarios are automatically derived from Operational Design Domain definitions and requirements, ensuring systematic coverage from the start.
- AI-driven automation: A fine-tuned LLM generates and parameterizes large-scale, simulation-ready scenarios across environments, actors, and edge cases.
- Built-in traceability: Every scenario is linked to requirements and ODD elements, enabling validation, auditability, and certification readiness.
- Continuous coverage & updates: Scenario libraries evolve dynamically as requirements and standards change, keeping validation up to date.
- Handles long-tail complexity: Agentic AI efficiently
creates rare and complex edge-case scenarios that are difficult to capture manually.
