To main content

The Edge AI Systems Reference Architecture - Orchestrating Autonomous and AI-Defined Systems Through GenAI and Agentic AI in the Intelligence Mesh

Abstract

The presentation outlines the evolving reference architecture needed to support the next generation of edge AI, moving beyond simple data processing to complex edge reasoning. It explores the convergence of novel heterogeneous solutions that optimise the hardware-software-AI-data quad required to run GenAI, agentic AI, Small Language Models (SLMs) and Vision Language Models (VLMs) efficiently at the edge. By defining a holistic edge AI technology stack that integrates trustworthy edge AI, software platforms and datasets, with hardware-aware optimisation, it is possible to lay the groundwork for mesh intelligence where edge systems possess the agency to sense, reason, plan, learn, memorise and execute tasks autonomously. These fundamental elements support the orchestration of the interactions within a collaborative mesh of autonomous AI-defined entities and components. The presentation opens the floor for discussion to provide a roadmap for shifting edge AI systems to dynamic, GenAI-enabled architectures capable of supporting the high computational demands of real-time agentic reasoning, with applications in AI-Defined Vehicles and robotics.

Category

Conference lecture

Language

English

Author(s)

Affiliation

  • SINTEF Digital / Sustainable Communication Technologies

Presented at

HiPEAC 2026 Conference

Place

Kraków

Date

26.01.2026 - 28.01.2026

Organizer

HiPEAC

Date

27.01.2026

Year

2026

View this publication at Norwegian Research Information Repository