What happens when ultra-low-power connectivity converges with high-performance machine learning right on the device? In this episode of the IoT Show, recorded live at IoT Stars, we sit down with Alexis, Chief Product Officer at Edge Impulse. Alexis shares insights from his journey from cloud environments and ultra-low bandwidth systems to leading product strategy at Edge Impulse, particularly following their acquisition by Qualcomm. We discuss how Qualcomm’s developer-first strategy keeps Edge Impulse independent while granting unprecedented access to bleeding-edge chipsets and ecosystem collaborations with partners like Arduino. Alexis also addresses the critical "walls" facing the industry—highlighting security vulnerabilities and the absolute necessity of responsible AI as development accelerates. Finally, we look at the massive evolution from Visual Language Models (VLMs) to Vision Language Action Models (VLAs), charting the clear path forward for robotics, physical AI, and intelligent edge systems that sense, think, and act safely. 🗒️ Key Topics Covered: - Qualcomm & Edge Impulse Integration: How the corporate integration maintains Edge Impulse's core DNA while leveraging massive silicon and engineering resources. - The Major Risk in Rapid Edge AI Adoption: Moving past hype to avoid historical IoT pitfalls like early IP camera hacks by focusing heavily on responsible AI and developer guidance. - The Changing Developer Landscape: Bridging the gap between traditional embedded hardware developers and incoming data scientists/machine learning engineers. - The Evolution of Edge Models: How cascaded models allow low-power Arduino devices to combine vision and small language models for rich local context. - The Rise of Physical AI (VLAs): Why Vision Language Action Models are revolutionizing robotics and enabling safe, predictable, and highly tested machine execution. Chapters: 0:00 - Welcome back to the IoT Show Live from IoT Stars 0:13 - Introducing Alexis, Chief Product Officer at Edge Impulse 0:38 - Qualcomm’s Developer-First Strategy & Edge Impulse Integration 1:34 - Transitioning from Low-Bandwidth Connectivity to Edge AI 2:19 - What Can Go Wrong? Security Risks and Responsible AI 3:13 - The Evolving Developer Community: Embedded vs. Data Science 4:14 - Beyond LLMs: Visual Language Models (VLMs) on Constrained Hardware 4:56 - The Future of Robotics: Vision Language Action Models (VLAs) 5:30 - Outro & What’s Next on the IoT Show #EdgeAI #EdgeImpulse #Qualcomm #PhysicalAI #EmbeddedSystems #VLA #IoTShow #TinyML #Robotics

Are we finally past the "IoT Hype" cycle? Recorded live at **IoT Stars during Embedded World 2026**, this episode of the IoT Show dives into the current reality of the industry. We sit down with industry veteran Ken Yu to discuss why 2026 is seeing a massive uptick in massive-scale deployments and how the developer's role is evolving. We move beyond simple data collection to discuss the necessity of edge AI in transforming raw sensor data into meaningful industrial actions. If you are a developer struggling to demonstrate the value of your IoT solution, this conversation explores why "dashboards and numbers" are no longer enough. 🗒️ Key Topics Covered: - The 2026 Momentum: Why massive deployments (reaching millions of devices) are finally becoming a reality in sectors like livestock monitoring and emergency safety. - The Edge AI Value Add: How AI is being used to analyze data locally, providing the "insightful actions" that industries actually need. - Hardware vs. Software: Navigating the unique challenges of hardware validation and changing customer behavior in the field. - Build vs. Buy: When developers should build custom hardware and when to leverage third-party solutions to solve common use cases. 🎙️ Live from Nuremberg: This conversation was captured during the IoT Stars event, bringing you the latest energy and insights directly from one of the industry's premier gatherings for IoT innovators. Chapters: 0:00 - Live from Embedded World 2026 0:26 - Introduction: Ken Yu and the 2026 IoT Landscape 1:10 - Is IoT Finally Taking Off? Current Momentum and Scale 1:46 - The Hardware Challenge: Why it’s harder than software 2:06 - Why Data is Useless Without AI for Analysis 2:54 - Build vs. Aggregate: Strategies for Solution Builders 4:00 - The Future of Edge AI and Industry Trends 5:15 - Closing Thoughts from IoT Stars

Are developers getting caught up in the AI hype, or is there a real-world path to running intelligence on constrained, battery-powered devices? In this episode of the IoT Show, recorded live at **IoT Stars during Embedded World 2026**, we sit down with Wienke Giezeman, CEO and co-founder of The Things Industries. Wienke shares invaluable, unfiltered perspectives gained over 11 years in the developer ecosystem, moving past investor buzzwords to look at what enterprises actually need. We dive into the technical intersection of LoRaWAN and TinyML/Edge AI. While you can't push massive model updates over low-bandwidth networks, low-power silicon innovations are opening doors for local inference (like audio or basic image processing) directly at the ultra-edge—allowing battery-operated devices to last for years while sending smart, event-driven notifications. Finally, Wienke gives us a teaser of his brutally honest IoT Stars talk on the grit required to build a scaling SaaS company by ignoring vanity metrics and listening to the raw truth from your users. 🗒️ Key Topics Covered: - LoRaWAN vs. 5G/Cellular: Why enterprise IoT thrives on long-distance, low-power networks when concrete walls and total operation costs break cellular options. - Edge AI on Tiny Devices: The rise of low-power MCU AI chips (Inatera, Brainchip, Alif, Efficient Computer) capable of running local inference without draining batteries. - Bandwidth Realities: Navigating the technical boundaries—why over-the-air ML model updates are too heavy for LoRaWAN, and why a "mature enough" fixed model is often all you need. - The "Hype Spectrum": Balancing bleeding-edge technology with legacy enterprise clients who still need foundational logic before adopting AI. - Grit in the IoT Market: Lessons from a 10-year journey focused strictly on user utility rather than investor expectations. Chapters: 0:00 - Live from IoT Stars Nuremberg 2026 0:26 - Introduction: Wienke Giezeman and The Things Industries 1:20 - The Enterprise Reality: LoRaWAN vs. Ubiquitous 5G 2:43 - Low-Cost Scaling: Lowering the Barrier to Industrialized IoT 3:04 - Private Networks vs. Public Infrastructure 3:35 - Can Ultra-Low Power Devices Handle Edge AI? 4:07 - Meeting the Enterprise Where They Are (Moving Past the Hype) 4:56 - Smart Filtering: Audio and Image Recognition on the Battery Edge 5:40 - Teaser: "Don't Piss on Me and Tell Me It's Raining" (Lessons on Grit) 7:15 - Market Momentum: Steady Growth vs. Shiny Objects

In this episode of The IoT Show, we dive deep into the world of Vision AI at the Edge with Muid Mufti from Au-Zone Technologies. If you’ve struggled with the complexity of creating, training, and deploying high-performance models for edge devices, this deep dive into EdgeFirst AI is for you. We explore the emerging concept of Spatial Perception—going beyond simple object detection to understand the environment in 3D space—and how Au-Zone is making this accessible for industrial and commercial IoT applications. 🔍 What’s inside this episode: - The Challenges of Edge AI: Muid breaks down the biggest hurdles developers face when moving Vision AI from the cloud to constrained edge hardware. - Introduction to EdgeFirst AI: Learn how this full-stack solution simplifies the entire development lifecycle. - Demo: EdgeFirst Studio: Watch a hands-on walkthrough featuring AI-powered automation for annotating massive image datasets—saving you weeks of manual labor. - Optimization & Deployment: Discover how to squeeze maximum performance out of your hardware without sacrificing accuracy. Whether you are an IoT architect, a data scientist, or a developer looking to integrate advanced computer vision into your next project, this episode provides a roadmap for building smarter, faster edge solutions. 🚀 Key Takeaways: - Understand the shift from traditional Vision AI to Spatial Perception. - Learn how to automate the most tedious parts of the AI pipeline. - Explore the tools needed to optimize models for real-time edge performance. 🔗 Resources & Links: Learn more about Au-Zone Technologies: https://www.au-zone.com/ Explore EdgeFirst Studio: https://edgefirst.studio/ 📚 Chapters 00:00 Vision AI at Edge 03:15 Edge AI Importance 06:49 Specialized Vision AI ️ 09:36 Edge AI Challenges 11:52 EdgeFirst AI Platform 14:32 Coffee Cup Detection 16:58 AI-Assisted Annotation 20:19 Model Training Session 25:28 Live Model Performance 28:37 Timing Analysis ⏱️ 31:22 Coffee Cup Training 34:40 Real-Life Scenarios 37:22 Ocean Cleanup Model 40:24 Future Discussions Powered by https://timeskip.io/?ref=olivier