EXPERT WEBINAR SERIES
Scaling scenario variations with neural reconstruction and world foundation models for AV training and validation
ON DEMAND
Gianmarco Macaro discusses how the Foretellix Foretify toolchain integrates with CARLA, NVIDIA Cosmos world foundation models (WFMs), NVIDIA NuRec, and Parallel Domain Replica to generate and vary synthetic and reconstructed autonomous vehicle (AV) scenarios.
Learn how scenario-based automation, neural reconstruction, and WFMs enable scaling validation, training, and support AV developers.
Key topics and takeaways:
- Formalization from requirements: Translating natural language safety requirements into formalized scenarios using OpenSCENARIO DSL and Foretify V-Suite libraries
- Hybrid runtime execution: Running co-simulations with Foretify, CARLA, NVIDIA Cosmos, and neural rendering engines for physics-aware realism
- Neural reconstruction: Using NVIDIA NuRec and PD Replica to transform raw logs into simulation-ready, sensor-fidelity digital twins
- Scenario variation: Learn how to adjust trajectories, insert synthetic actors, use different behavioral models for agents, and generate “what-if” scenario variations at scale
- Constraint-driven scalability: Applying constraint solving to OSC DSL code – to produce thousands of valid, intent-preserving scenario variations
- Coverage & KPIs: Connecting reconstructed and synthetic scenarios to unified Foretify Evaluate dashboards for ODD coverage tracking and traceable safety evidence
The Presenter
Gianmarco Macaro – Application Engineering Director, Americas at Foretellix, leads the Application Engineering team across North and South America. Prior to Foretellix, Gianmarco held engineering and sales roles at Humanetics Innovative Solutions, focusing on ADAS and AV robotic testing solutions. He holds a Master’s and Bachelor’s degree in Mechanical Engineering from the University of Bologna