FEV has partnered with Microsoft to bring generative AI capabilities into vehicles using Nvidia GPU-accelerated computing and AI model microservices. The collaboration focuses on enabling multimodal interactions, including voice, text and gesture controls, directly inside the vehicle, reducing dependence on a continuous internet connection.

At the core of the solution are small language models (SLMs), including Microsoft's Phi-4-mini-instruct model, running on Nvidia Drive AGX accelerated computing platforms. The technology allows drivers to configure vehicle functions, such as dashboard settings and personalized vehicle profiles, through natural-language voice commands. It also serves as a local fallback intelligence layer for cloud-based large language models (LLMs).

By processing AI inference within the vehicle, embedded SLMs enhance the responsiveness, reliability and availability of AI-powered functions, even in areas with limited or no network connectivity. The approach also helps reduce backend and cloud infrastructure costs by enabling certain functions to be performed locally rather than relying exclusively on cloud-based LLMs. FEV said this could help automakers scale software-defined vehicle (SDV) features more cost-effectively.

“Our collaboration with Microsoft and Nvidia showcases how small, efficient language models can transform in-vehicle experiences, delivering powerful functionality without the overhead of larger systems,” said Thomas Hülshorst, group vice-president of Intelligent Mobility and Software at FEV.

This content may be AI-assisted and is composed, reviewed, edited and approved by Mobility Global.