We’re excited to announce that Embedl will exhibit and present at InCabin USA 2025, which will take place June 10–12 at Huntington Place in Detroit. This year’s event brings together industry leaders to explore the latest innovations in automotive in-cabin technology, and we're proud to contribute to the conversation.
Our very own Peter Kristiansen, Head of Business Development, will deliver a talk titled: "Why, How, and Where to Deploy Generative AI and Other State-of-the-Art AI to Edge Devices"
Why This Matters Now
Generative AI—especially Large Language Models (LLMs) and Transformers—has rapidly become the engine behind today’s most advanced intelligent systems. From conversational assistants to real-time decision-making in autonomous vehicles, these models are reshaping how we interact with technology.
However, running these models in the cloud introduces major bottlenecks:
- Privacy concerns
- Dependence on connectivity
- High operational costs
This has created a pressing need to move AI processing to the edge—in vehicles, on embedded devices, and within constrained environments where speed, privacy, and independence are crucial.
Inside the Talk: What You’ll Learn
Peter’s presentation will explore practical strategies for deploying state-of-the-art AI models on edge devices without sacrificing performance or reliability. Highlights include:
-
A deep dive into System-on-Chip (SoC) platforms from Texas Instruments, Nvidia, Qualcomm, and Ambarella, and how they handle complex model inference.
-
A review of model compression techniques—including quantization, pruning, and knowledge distillation—and the performance trade-offs that come with each.
-
There is an emerging focus on Small Language Models (SLMs), which are designed to offer near-LLM capabilities with dramatically reduced hardware requirements.
Peter will also present real-world results from our collaboration with an OEM, where customized Llama 3.2-based SLMs were distilled and fine-tuned for automotive edge use cases. These experiments show that it’s not only possible—but practical—to embed intelligent, responsive AI in today’s vehicles.
SLMs: The Future of On-Device AI
A key trend Peter will cover is the rise of SLMs (1–10 billion parameters)—models that balance the power of LLMs with the performance demands of edge deployment. With organizations like Meta leading the way (as seen in the September 2024 release of Llama 3.2), the future of edge AI is increasingly defined by compact, efficient, and domain-specific models.
But challenges remain. Even advanced SoCs like Qualcomm’s Snapdragon need engineering workarounds to support anything beyond pre-packaged models. Peter will shed light on how developers can navigate this landscape with smart model design and optimization strategies.
Join Us at InCabin USA
Generative AI is no longer confined to the cloud. The edge is where innovation meets reality—and Embedl is proud to be driving that shift forward.
Visit us at our booth or schedule a meeting here: https://www.embedl.com/schedule-a-meeting
Get your tickets to InCabin USA 2025 here: https://incabin.com/usa/
Stay tuned and meet us at InCabin USA 2025!