[ivory-search id="3463" title="Default Search Form"]
Search
Register here
Search
Type below to search:
Room S
Workshop
In person
LeadersGoldDiscovery

Intelligence at the edge

    • Date
    10 July 2026
    Timeframe
    09:00 - 12:15 CEST
    Duration
    3h 15 minutes
    • Days
      Hours
      Min
      Sec
    At the breaking edge of innovation, Edge AI is ushering in a new era of data processing – enabling real-time decision-making without high-bandwidth cloud connectivity.
     
    This expert workshop at the 2026 AI for Good Global Summit explores the technical requirements for deploying intelligence directly on local hardware which is vital for mission-critical tasks and resource-constrained environments. By processing data at the source, Edge AI enables lower latency, energy efficiency, and offline functionality, while prioritizing data sovereignty and privacy. Find out how these decentralized systems provide the reliability needed to address global challenges where traditional infrastructure is absent.
     
    Moving from theory to implementation, the workshop covers practical methods for developing lightweight, sustainable AI models that minimize environmental footprints. Key topics include federated learning, explainability at the edge, and strategies for scaling from prototypes to production. The session will highlight open-source tools and educational materials that support a trustworthy, interoperable ecosystem for Edge AI solutions.
     
    Objectives The main objectives of the workshop are to:

    1. Discuss the hardware and software requirements for deploying lightweight ML models on local, resource-constrained devices.
    2. Explore how Edge AI minimizes environmental footprints by reducing energy consumption and data transmission.
    3. Identify strategies to deploy intelligence in “offline” environments, ensuring AI benefits regions with limited connectivity (LDCs/SIDS).
    4. Highlight open-source tools, datasets, and educational initiatives (like TinyML4D) that foster local capacity building.
    Schedule

    In many real-world scenarios, cloud-centric AI systems are not viable due to latency, bandwidth, cost, or infrastructure constraints.

    This makes it necessary to rethink how intelligence is executed and how data flows across the system, moving processing closer to where data is generated. In this context, the combination of edge computing and connectivity enables a new class of distributed intelligence, where processing and communication are tightly integrated.

    This keynote will explore how Edge AI can support faster decision-making, lower latency, improved privacy, and greater operational resilience by processing data locally while remaining connected to the broader IoT ecosystem.

    Speaker(s)

    The TinyML4D Academic Network connects more than 50 universities across Africa, Latin America, and Asia around a shared conviction: that the people closest to local problems should be the ones building the technology to solve them. Its focus is machine learning on low-cost, low-power devices — putting intelligence where data is generated, in farms, clinics, and field stations far from reliable connectivity. Over the past years, the network has trained educators and students, developed open educational materials, and supported locally driven projects in agriculture, environmental monitoring, and health, building institutional capacity that stays and grows within each region. Today it is broadening from TinyML toward the wider field of EdgeAI. This talk shares what the network has learned about capacity building, and why a distributed academic community is one of the most effective ways to make EdgeAI genuinely inclusive.

    Speaker(s)

    Data selection molds an AI system's world view. This data dependency is particularly acute in resource-constrained, context-specific edge environments. Yet, data curation has historically been under-valued, with computer vision applications still relying on ethically problematic datasets. This talk introduces the Fair Human-Centric Image Benchmark (FHIBE), a consensually-sourced dataset for evaluating and comparing bias across various computer vision tasks common in Edge AI use cases. In the talk I will highlight key attributes of the dataset, share lessons learnt during its curation, and opportunities for improving the robustness of Edge AI deployments through bias evaluations.

    Speaker(s)

    Reliable internet connectivity cannot be assumed in many parts of the world, making offline AI essential for education, healthcare, and access to knowledge. This joint presentation brings together the GSMA African AI Language Initiative and Libraries Without Borders (Bibliothèques Sans Frontières) to demonstrate how AI can be deployed entirely on low-cost edge devices without cloud connectivity. We present a voice-first offline AI platform built on small language models, speech recognition, and reproducible edge hardware, alongside benchmark results on Raspberry Pi 5 and NVIDIA Jetson Orin NX that highlight the performance of quantized models in resource-constrained environments. We also showcase real-world deployments supporting teachers, community health workers, students, and local language communities through offline AI assistants, local knowledge bases, and community-driven datasets. Together, these experiences demonstrate how open, offline AI can expand access to trustworthy AI while supporting local languages, data sovereignty, and digital inclusion.

    Speaker(s)
    Share this session with your network
    • Session starts in
      Days
      Hours
      Min
      Sec
    Add to Calendar
    Related stories

    Frontier Research, Real-World Impact: Technology Innovation Institute (TII) at the...

    Returning to Geneva, TII showcases live demonstrations that turn frontier research into capability the world can use by Dr. Najwa...
    Global Summit

    AI and frontier technologies take centre stage at ITU Kaleidoscope...

    As artificial intelligence and other frontier technologies continue to redefine the boundaries of innovation, Kaleidoscope 2026: AI and frontier technologies...
    Global Summit

    Who controls frontier AI? Building a plan for diffused access

    In 2026, more than a billion people use a generative AI tool every month. That figure sounds like momentum –...
    Global Summit

    Are you sure you want to remove this speaker?