Speaker biography
Stephen Lernout is the co-founder and CEO of Deontic, a company building safe-autonomy tooling for physical AI and automated driving. A serial entrepreneur with a background in technology sales and marketing, he previously led strategy and innovation at language-technology ventures including Miia and Nalantis.
Presentation
As automated-driving programmes expand across regions, vehicle platforms and software releases, validation teams face a growing challenge. The primary bottleneck is no longer simulation capacity alone, but the ability to systematically translate Operational Design Domain constraints, engineering requirements, safety objectives and regulatory expectations into complete, testable and traceable scenario coverage. As the ODD approaches the open world, the challenge is not simply producing more scenarios, but understanding the quality of the resulting coverage and proving it with defensible evidence. This work presents Deontic's prompt-driven, ODD-first approach to scenario engineering and validation-evidence generation. The framework ingests ODD definitions, technical requirements and compliance inputs, then uses agentic workflows to author, parameterise and manage simulation-ready scenario libraries. The generated scenarios cover road geometry, traffic infrastructure, actor behaviour, weather, visibility, work zones and long-tail interactions. The approach also produces requirement-aligned parameter sets, coverage plans, test matrices and provenance metadata, maintaining traceability from each ODD condition and requirement to scenario intent, simulation execution, coverage results and validation evidence. Rather than replacing existing simulation or validation infrastructure, Deontic operates as an intelligence and API layer across the engineering toolchain. It feeds simulators, test-management platforms, continuous-integration pipelines and evidence-management systems with structured scenarios and defensible test logic. This strengthens the upstream decisions around what should be tested, why it should be tested, how sufficient the coverage is, and how the resulting evidence should be organised, reused and maintained. The approach delivers value in three areas. First, it improves scalability by accelerating scenario development for new ODDs, routes, vehicle configurations and markets. Second, it strengthens governance by maintaining compliance-aligned documentation and audit-ready evidence as requirements and standards evolve. Third, it improves release efficiency by reducing manual rework, broadening edge-case coverage and synchronising requirements, scenarios, tests and results throughout the validation lifecycle. IPG Automotive contributes as a validation-technology and ecosystem partner, enabling generated scenarios and test plans to be managed and executed through VIRTO and IPG CarMaker. This collaboration demonstrates how ODD-first scenario intelligence can integrate with established virtual-validation environments while remaining applicable across broader simulation and validation toolchains. The result is a shift from simulation enablement towards validation assurance. Organisations can demonstrate not only that sufficient testing was performed, but that the right scenarios were selected, coverage quality was assessed against defensible criteria, and the resulting evidence remains traceable, reusable and ready for audit.