Foretellix, a provider of physical AI safety and data infrastructure, announced a reference solution for the NVIDIA Alpamayo ecosystem. Foretellix enables developers to build AI-powered self-driving systems to accelerate data curation, synthetic data generation (SDG), testing, and validation workflows.
Foretellix reference solutions provide the guidance, methodologies, and tools needed to train and validate AV stacks with higher reliability and scale for AI-powered self-driving development environments.
“The transition to AI-driven autonomous driving will fundamentally change the way self-driving vehicle systems are developed and validated,” said Ziv Binyamini, CEO and co-founder of Foretellix. “Developers can no longer rely on traditional software verification approaches. Our NVIDIA Alpamayo-based reference solution demonstrates the Foretellix data-centric infrastructure needed to train, validate, and securely scale next-generation AI-driven systems.”
Also read: AIThority interview with Rohit Agarwal, Founder and CEO of Portkey
Data-centric infrastructure for physical AI
The workflow begins with denoising the AV drive log. This is the basis for extracting ground truth from data. This process enables both the labeling of temporary scenarios and the design of synthetic scenarios, allowing developers to precisely replicate specific scenarios or create controlled variations of them.
This fundamental step is followed by data curation and warehouse exploration, where developers identify high-value driving segments that serve as a baseline for synthetic data generation. V&V engineers can instantly gain visibility into specific operational design domain (ODD) areas that the system has encountered or tested in both virtual and real-world environments. Foretellix solutions organize this data into structured warehouses rather than unorganized data lakes.
Bridging the ODD gap with synthetic data generation
Behavioral ODD coverage analysis and SDGs are at the heart of the physics AI toolchain. The toolchain identifies specific gaps in the ODD and allows test engineers to design and edit new synthesis scenarios that directly fill the coverage gaps.
Secure and scalable
Through Foretellix Foretify Scenario Designer integrated with NVIDIA Omniverse NuRec, engineers can modify actor behavior, add new actors to the reconstructed scene, and analyze ODD integrity. These scenes can be further enhanced and expanded using NVIDIA Cosmos. Generating diverse scenarios at scale is key to validating whether a stack is ready to be safely deployed in complex physical environments.
Foretellix will be demonstrating its NVIDIA Alpamayo-based reference solution at CVPR on June 5th at 3:40pm at the NVIDIA Expo Theater, in booth 826 during the show.
