Cerence Pioneers Automotive-Specific LLM in Collaboration with NVIDIA, Powering the Future of In-Car Experiences

Cerence has created an LLM-based platform leveraging its extensive automotive dataset and tech stack to deliver enhanced experiences for end users

BURLINGTON, Mass., Dec. 19, 2023 (GLOBE NEWSWIRE) -- Cerence Inc. (NASDAQ: CRNC), AI for a world in motion, today introduced a pioneering, automotive-specific large language model, CaLLM, powered by NVIDIA technology. CaLLM (Cerence Automotive Large Language Model) serves as the foundation for Cerence’s next-generation in-car computing platform, running on the NVIDIA DRIVE platform. Cerence is working with NVIDIA to solve several key automaker challenges: accelerating time to market by enabling new user experiences to be deployed via a cloud integration with both existing embedded systems and new generative AI-powered domains, giving OEMs an advantage as drivers demand increasingly intelligent features and capabilities inside the car.

As automakers and mobility OEMs look to leverage generative AI and LLMs to improve the user experience, CaLLM meets several critical needs. It has a unique level of automotive-specific intelligence, leveraging Cerence’s extensive automotive expertise and fine-tuned, growing automotive dataset encompassing billions of tokens to deliver integrated in-car user experiences beyond those of general knowledge LLMs. CaLLM is distinctly able to support automotive functions, features, and requirements and is deeply customizable for automakers through training, fine-tuning, and bespoke applications. Its features include user personalization as well as built-in information retrieval for Cerence’s generative AI-powered applications like Cerence Car Knowledge.

Cerence is building CaLLM with the NVIDIA AI foundry service, which includes NVIDIA AI Foundation Models, NVIDIA AI Enterprise software, and NVIDIA accelerated computing. Cerence will first train CaLLM on its extensive dataset leveraging NVIDIA DGX Cloud and NVIDIA DGX systems and then develop the capabilities required for in-car user experiences. Second, to help achieve an ultrafast user experience, Cerence will deploy CaLLM in an NVIDIA-accelerated infrastructure with NVIDIA AI Enterprise, an end-to-end, cloud-native, secure software platform that accelerates training, fine-tuning, and inferencing of LLMs with the NVIDIA NeMo framework and the NVIDIA TensorRT-LLM optimization library.