Autobrains is expanding its collaboration with NVIDIA to bring advanced AI reasoning into real-world autonomous driving systems. The Tel Aviv-based company is using NVIDIA’s Alpamayo 2 Super, a 34-billion-parameter foundation model, to develop specialised Driving Agents designed to operate in production vehicles in real time.
The approach uses Agentic AI to convert the large foundation model into smaller, task-specific agents. An orchestrator activates only the agents needed for a particular driving situation, reducing onboard computing requirements without requiring the full model to run inside the vehicle.
AI Training Targets Complex Driving Scenarios
Autobrains is also using Alpamayo’s reasoning capabilities to generate rare and complex driving scenarios that could take millions of road miles to encounter naturally.
These scenarios help train the specialised agents and identify weaknesses before deployment, supporting the development of more capable autonomous driving systems.
Level 4 Programmes Expand
Autobrains and NVIDIA are already collaborating on Level 4 autonomous driving programmes with VinFast in Southeast Asia and Uber in Munich.
NVIDIA said AI reasoning could play an important role in advancing autonomous vehicle technology.
“Reasoning AI is the next great leap in autonomous driving,” said Rishi Dhall, NVIDIA’s vice president of automotive.
The collaboration combines large-scale AI reasoning with specialised onboard systems, aiming to make advanced autonomous driving capabilities more practical for real-world vehicle deployment.