Revolutionizing Embedded Computer Vision With Cutting-Edge AI Processors EDGE AI POD podcast

42:43

Send us a text Support the show Learn more about the EDGE AI FOUNDATION - edgeaifoundation.org

1
The Pipeline Is Stalling: America's Declining Innovation Edge 13:41

9 days ago13:41

13:41

America's innovation pipeline stands at a dangerous crossroads. Federal research funding has dropped 10% in real terms over the past decade, traditional public-private collaboration models are fragmenting, and barriers to international talent continue to rise. These challenges threaten the three-pillar foundation that has powered American technological leadership for generations. Our conversation dives deep into Harvard University Professor Vijay Janapa Reddi's compelling analysis of this critical situation. The statistics are striking: researchers have lost 26% of their purchasing power since 2014, forcing tough choices about graduate positions, equipment purchases, and research directions. Meanwhile, 79% of computer science graduate students are international, underscoring our reliance on global talent. When 55% of billion-dollar American startups have immigrant founders, restrictive immigration policies amount to what experts call "national self-sabotage." The impacts extend far beyond elite institutions. America's innovation ecosystem encompasses land-grant universities, HBCUs, community colleges, and state flagships - all dependent on stable federal support. The Edge AI Foundation, where Professor Reddy serves on the board, exemplifies one promising response: creating structured collaboration between universities and industry on emerging technologies like neuromorphic computing and edge-based AI. This approach helps bridge the crucial gap between academic research and commercial application. Revitalizing our innovation ecosystem demands a coordinated strategy: expanded industry consortia with federal matching funds, innovation co-labs on university campuses, dedicated STEM green cards, streamlined visa processes, and predictable research funding increases. The stakes couldn't be higher - our economic future and technological leadership hang in the balance. How might you contribute to rebuilding America's innovation pipeline? Explore more through the People's Pledge for American Higher Education and join the conversation about securing our innovative future. ➡️ Visit HERE to read The Pipeline Is Stalling whitepaper ➡️ Visit HERE to sign the pledge Send us a text Support the show Learn more about the EDGE AI FOUNDATION - edgeaifoundation.org…

1
Beyond the Edge - Analyst Insights at AUSTIN 2025 59:38

16 days ago59:38

59:38

What happens when four tech industry veterans sit down to dissect the future of Edge AI? A no-holds-barred conversation that cuts through the hype and delivers actionable insights for anyone working in this rapidly evolving field. The panelists tackle the burning question head-on: How can Edge AI avoid becoming the "box of cables in your garage" that IoT transformed into? Through candid discussion, they reveal that success depends less on technical prowess and more on solving real business problems with clear ROI. As one panelist bluntly states, "The customer doesn't care about MQTT or AMQP protocols—nobody cared. They care about outcomes." The conversation weaves through crucial territory—the boundary between IT and OT systems, the evolving relationship between cloud and edge architectures, and the stark reality of skills shortages in the semiconductor industry. With semiconductor companies doubling in size over the next eight years while facing an aging workforce, the panel highlights urgent needs for reskilling and education initiatives. Perhaps most provocatively, the panel addresses the uncomfortable truth about AI adoption: "How many employees is your technology going to replace?" This question, frequently asked by VCs, underscores the economic drivers pushing Edge AI forward in an era of changing workforce demographics and productivity challenges. For small and medium businesses, there's both opportunity and risk as AI becomes embedded in SaaS offerings and purpose-built solutions emerge for specific industry problems. The panelists predict that Edge AI might actually benefit smaller players more than large enterprises by democratizing access to sophisticated capabilities through platform-based approaches. Whether you're developing Edge AI solutions, investing in the space, or considering adoption for your business, this discussion provides essential context for navigating the significant opportunities and challenges ahead. Subscribe to hear more industry experts cut through the noise and deliver practical wisdom for the intelligent edge. Send us a text Support the show Learn more about the EDGE AI FOUNDATION - edgeaifoundation.org…

1
Dan Cooley of Silicon Labs - The 30 Billion Dollar Question: Can AI Truly Live on the Edge? 23:59

23 days ago23:59

23:59

Imagine a world where your smart glasses don't just identify objects but tell stories about what they see—all while running on a tiny battery without heating up. This cutting-edge vision is becoming reality as semiconductor companies tackle the monumental challenge of bringing generative AI capabilities from massive cloud data centers down to microcontroller-sized devices. The semiconductor industry stands at a fascinating crossroads where artificial intelligence capabilities are pushing beyond traditional cloud environments into battery-powered edge devices. As our podcast guest explains, this transition faces substantial hurdles: while cloud-based models expand from millions to trillions of parameters, embedded systems must dramatically reduce their footprint from terabytes to gigabytes while still delivering meaningful AI functionality. With projections showing IoT devices consuming over 30 terabit hours of power by 2030 and generating 300 zettabytes of data, the need for local processing has never been more urgent. For developers creating wearable technology like smart eyewear, constraints become particularly challenging. Weight distribution, battery life, and computing power must all be carefully balanced while maintaining comfort and style. The hardware architecture required for these applications demands innovative approaches: shared bus fabrics that enable different execution environments, strategic power management that activates high-performance cores only when needed, and neural processing units capable of handling transformer operations for generative AI workloads. Most impressively, current implementations demonstrate YOLO object detection running at just 60 milliamps—easily within battery operation parameters. The $30 billion embedded AI market represents a tremendous opportunity for innovation, but also requires robust software ecosystems that help traditional microcontroller customers without AI expertise navigate this complex landscape. As next-generation devices begin supporting generative capabilities alongside traditional CNN and RNN networks, we're witnessing the dawn of truly seamless human-machine interfaces. Ready to explore how these technologies might transform your industry? Listen now to understand the future of computing at the edge. Send us a text Support the show Learn more about the EDGE AI FOUNDATION - edgeaifoundation.org…

1
The Future of Domain-Specific AI Search Lies in Targeted Agent Systems 1:00:46

30 days ago1:00:46

1:00:46

Imagine your edge device having the ability to search for exactly what you need, exactly when you need it, without hallucinations or irrelevant information. That's the promise of Snipe Search's agent orchestration system, presented by co-founder Wassim Kezai in this eye-opening EDGE AI TALKS session. Most organizations struggle when implementing RAG systems with their corporate data. The truth is, unstructured corporate knowledge is often messy and inconsistent, leading to unreliable AI responses. Semantic matching issues in traditional retrieval systems further compound these problems, especially when deployed at the edge where specific, accurate information is crucial. Wasim unveils an innovative approach that deploys specialized AI "detective" agents to search for information from authoritative sources. Unlike brute force search methods, these agents intelligently target reliable information based on hierarchical importance. Web agents crawl and cross-reference websites, image agents find relevant visuals, scholar agents specialize in academic information, and video agents can even pinpoint the exact timestamp in video content that answers your query. What sets this approach apart is its adaptability to domain-specific knowledge and verification frameworks. Companies can customize how information is validated based on their standards, ensuring relevance and accuracy. While traditional RAG systems respond in seconds, Snipe Search's 30-second average response time delivers significantly higher quality information – a worthwhile trade-off for mission-critical applications. The platform integrates easily with any LLM or chatbot through Docker, API, or direct integration, making it accessible for organizations of all sizes. As edge computing continues to grow, having efficient, accurate search capabilities becomes increasingly important for reducing cloud dependencies, enhancing privacy, and delivering better user experiences. Ready to transform how your edge devices access and utilize knowledge? Explore Snipe Search's platform launching in the coming weeks and discover how intelligent search can enhance your edge AI deployments. Send us a text Support the show Learn more about the EDGE AI FOUNDATION - edgeaifoundation.org…

1
Cleaning Our Oceans with Edge AI 58:47

5 weeks ago58:47

58:47

The crisis of ocean plastic pollution demands innovative solutions, and Brad's team at Ozone Technologies is answering the call with cutting-edge AI technology. Their collaboration with The Ocean Cleanup has yielded a remarkable system that's transforming how we detect and map marine debris. At the heart of this solution is ADIS (Autonomous Debris Imaging System) – a compact, rugged camera powered by edge AI that mounts to merchant vessels traveling across our oceans. Using sophisticated computer vision models running on NXP i.MX 8M Plus processors, these devices scan the water's surface, distinguishing tiny pieces of floating plastic from wave crests even in challenging marine conditions. What makes this approach revolutionary is its scalability. Rather than requiring dedicated research vessels, ADIS piggybacks on existing shipping routes, creating an expanding network of detection points across global waters. The system operates completely autonomously, requiring minimal power and no continuous internet connection. It stores detection data onboard, uploading it only when ships return to port. The technical challenges overcome by Brad's team are impressive – from waterproofing the hardware to withstand immersion and pressure spray to developing specialized tracking algorithms that can differentiate persistent floating debris from transient wave patterns. Perhaps most remarkable is how they've simplified deployment with fool-proof installation instructions that require no technical expertise from ship crews. This mapping data serves a critical purpose, helping The Ocean Cleanup optimize their System 3 recovery operations in the Great Pacific Garbage Patch. Their massive 2.2km floating barrier (nicknamed "Josh") funnels debris into collection zones, but knowing where to deploy is essential for efficiency. ADIS provides that critical intelligence. Beyond mapping, we're seeing the completion of a virtuous cycle as recovered plastic finds new life through creative recycling – including limited-edition Coldplay vinyl records made from ocean plastic. Want to see more innovations tackling our planet's pressing environmental challenges? Join us at the upcoming AJI Foundation Summit in Austin or explore edgifoundation.org/events for opportunities to connect with leaders in this space. Send us a text Support the show Learn more about the EDGE AI FOUNDATION - edgeaifoundation.org…

1
Smart Chips, Big Dreams: How NXP is Changing the AI Game 58:31

6 weeks ago58:31

58:31

The artificial intelligence landscape is undergoing a fundamental shift – moving away from simply compressing cloud models to fit edge devices toward developing AI that's truly "born at the edge." Davis Sawyer, AI Product Marketing Manager at NXP Semiconductors, guides us through this transformation and reveals how NXP is revolutionizing the way intelligence is built into our everyday devices. Sawyer begins by redefining what we mean by "the edge" – a vast spectrum ranging from network infrastructure handling millions of connections down to the dozens of microcontrollers in modern home appliances and vehicles. What makes edge AI unique isn't just about size constraints, but the fundamentally different operating environment. While cloud models enjoy virtually unlimited power and standardized computing environments, edge devices face strict limitations in form factor, power consumption, and thermal management. The heart of NXP's approach is their EIQ software stack – a comprehensive toolkit that spans their entire product range from low-power MCUs to high-performance MPUs. Two innovations stand out as particularly revolutionary: Time Series Studio brings AutoML capabilities to sensor data, enabling non-AI experts in manufacturing, energy, and other sectors to build powerful anomaly detection models without deep machine learning expertise. Meanwhile, their approach to generative AI uses "RAG on steroids" (Retrieval Augmented Generation) to create systems that are not only compact enough for edge deployment but also inherently more secure and private. The real-world impact is already evident in applications ranging from precision agriculture robots to healthcare systems that combine multimodal sensing for contact-free patient monitoring. Perhaps most impressive is the rapid pace of innovation – within just months, NXP's edge-optimized language models have seen response times drop from two seconds to less than half a second, making conversational interfaces truly viable on embedded devices. Looking ahead, Sawyer predicts we're moving toward a new era where edge AI becomes increasingly agentic – focusing not just on human-machine interfaces but on optimizing machine-to-machine workflows in factories, robotics, and automation. Join us to discover how the future of intelligence isn't trickling down from the cloud, but rising up from the edge where our data is born. Send us a text Support the show Learn more about the EDGE AI FOUNDATION - edgeaifoundation.org…

1
Counting What Counts: How Edge AI is Saving Japan's Seafood Industry 59:18

7 weeks ago59:18

59:18

When SoftBank's team was approached by Japanese fishermen struggling with declining catches and inefficient practices, they knew technology could offer a solution. Their response: an innovative Edge AI system that transforms traditional aquaculture through sophisticated computer vision running directly on smartphones. Japan's fishing challenges are emblematic of global food security concerns. With the country's self-sufficiency rate at just 38% and fish catches halved from peak levels, the stakes couldn't be higher. Traditional aquaculture has relied heavily on intuition rather than data, making it nearly impossible to optimize feeding (which represents 60-70% of operational costs) or accurately monitor fish populations. The SoftBank and AIZip collaboration tackles these challenges through a remarkable edge computing approach. What makes their solution particularly groundbreaking is how it functions in environments with zero infrastructure—no power supply, no connectivity, and corrosive saltwater everywhere. By developing density-based crowd counting AI models that run efficiently on edge devices, they've created a system that can count hundreds of fish with 96% accuracy in clear water and 86% accuracy in muddy conditions. Perhaps most fascinating is their development process. Unable to collect sufficient real-world training data underwater, the team developed a sophisticated Unity-based simulation that generates realistic fish behavior under various conditions. This simulator provided 65% of their training data, complemented by manual observation from divers who documented actual fish behavior at different depths and feeding stages. The result is an AI system that not only counts fish but can potentially detect hunger levels, health issues, and optimize feeding schedules. This CES 2024 Innovation Award-winning technology demonstrates how Edge AI can transform traditional industries without requiring massive infrastructure investments. By bringing intelligence directly to the point of data collection, even in the most challenging environments, we're witnessing the beginning of a new era in sustainable food production. Whether you're involved in agriculture, environmental monitoring, or any field requiring intelligent sensing in harsh conditions, the lessons from this underwater AI revolution could transform how you approach your next challenge. Send us a text Support the show Learn more about the EDGE AI FOUNDATION - edgeaifoundation.org…

1
Career EDGE: Navigating the Job Market in the Age of Edge AI 1:28:56

8 weeks ago1:28:56

1:28:56

From 5V Tech and the EDGE AI FOUNDATION, unlock the potential of Edge AI and transform your career path with insights from leaders in the field. Join our engaging conversation featuring Luke Perrins from 5V, Gregory from Infineon Technologies, and Professor Eiman Kanjo from Nottingham Trent University, as we uncover how Edge AI is revolutionizing industries with its unique ability to operate independently of cloud connectivity. Learn from seasoned professionals like Stephen Davis, an executive leadership coach, Kari, a corporate recruiter, and Martin MacDonald, COO at AI startup Weteeq, about the innovative applications and benefits of this technology, including improved latency and enhanced device autonomy. Explore the diverse and dynamic skill set essential for success in the Edge AI realm. Our experts discuss the importance of mastering embedded software, programming, and hardware understanding, while also emphasizing cross-disciplinary collaboration and soft skills like communication. Whether you're a graduate or a seasoned professional, discover actionable advice on crafting standout resumes and CVs tailored to this evolving field, as well as strategies to showcase your personal projects and achievements effectively. Prepare to stand out in your job applications and interviews with expert tips on leveraging networking opportunities and professional platforms like LinkedIn. Discover the critical role of personal branding and cultural understanding, and gain insights into future careers in Edge AI, where versatility and innovation are key. Tune in to this insightful episode packed with practical strategies to help you excel in the cutting-edge world of Edge AI. Send us a text Support the show Learn more about the EDGE AI FOUNDATION - edgeaifoundation.org…

1
Revolutionizing Wi-Fi Sensing with Machine Learning and Advanced Radio Frequency Techniques 13:24

9 weeks ago13:24

13:24

What if the future of Wi-Fi could pinpoint your location down to 30 centimeters? Join us as Joseph Chueh from National Tsinghua University unveils the astonishing potential of Wi-Fi sensing when integrated with machine learning. Joseph brings his wealth of experience in semiconductor research and business development to the table, discussing the revolutionary application of existing frequencies like 2.4G and 5G for tasks including human activity recognition and intruder detection. This episode unpacks how Channel State Information (CSI) is at the heart of extracting precise data for machine learning, while also addressing the technical hurdles of hardware optimization and interference management. Discover how increasing the degrees of freedom in Wi-Fi systems can be a game-changer for radio frequency technology. Joseph explains how adding more channels or phase coordination expands the sample space for channel information, paving the way for more efficient decision-making. We explore solutions like transmitter-side coding and the impact of transmission models like OFDM and OFDMA on Wi-Fi sensing capabilities. Joseph paints a vivid picture of a future where Wi-Fi sensing becomes not only more accurate but also more cost-effective and accessible, making it a promising feature in both today's Wi-Fi technologies and upcoming 6G systems. Whether for robotics or enhancing room-scale environments, the insights shared in this episode offer a glimpse into an exciting wireless frontier. Send us a text Support the show Learn more about the EDGE AI FOUNDATION - edgeaifoundation.org…

1
From Ideas to Reality with Edge AI Expertise from embedUR and ModelNova 59:56

10 weeks ago59:56

59:56

How can we revolutionize everyday objects into smarter, more responsive entities? Join us in this exciting episode as we explore the cutting-edge world of Edge AI, fresh from the vibrant showcases at CES 2025. We promise an in-depth look at the future of technology deployment with our special guests Eric and John from EmbedUR. With their extensive expertise in embedded products, they bring a fascinating perspective on the importance of blueprints for tech commercialization in today's rapidly evolving landscape. Our conversation with Eric Smiley and John Marconi introduces the innovative ModelNova, a transformative tool designed to speed up the journey from ideation to proof of concept for tiny edge AI devices. Discover how ModelNova's curated model zoo and datasets can empower developers, ensuring efficient performance even under resource constraints. We dive into practical insights like community contributions, adapting large AI models for small devices, and how tools like Model Nova democratize access across various chip platforms, turning ambitious ideas into reality. This episode doesn't just stop at the technical nuances; it goes further into the realm of edge AI product development. We share a compelling story of transforming a bicycle helmet camera with object detection to enhance rider safety, illustrating the complexities of selecting the right hardware and the critical role of blueprints in this journey. From MLOps integration to cloud connectivity for continuous updates, our discussion emphasizes collaboration within the tech ecosystem to tackle the challenges of AI deployment. Tune in to learn how these advancements are not only reshaping technology but also enriching everyday life by making objects around us smarter and more responsive. Send us a text Support the show Learn more about the EDGE AI FOUNDATION - edgeaifoundation.org…

1
Reimagining Edge AI using Breakthrough Tools with Ali Ors of NXP 16:00

11 weeks ago16:00

16:00

NXP Semiconductors is revolutionizing the integration of AI within embedded systems, focusing on enhanced tools and dedicated acceleration. Their latest innovations like the EIQ Software Stack and EIQ Time Series Studio are designed to streamline AI deployment, bridging the gap for developers transitioning into AI methodologies. • Overview of NXP’s focus on automotive, industrial, and smart home markets • Introduction of new AI-accelerated MCUs and application processors • The importance of effective data management and compute efficiency • Discussion of the EIQ Software Stack and EIQ Toolkit for developers • Insights into the EIQ Time Series Studio for automated model generation • Highlights of challenges faced by developers in AI implementation • Key takeaways on NXP's commitment to enhancing edge AI capabilities Send us a text Support the show Learn more about the EDGE AI FOUNDATION - edgeaifoundation.org…

1
Revolutionizing AI Deployment with Open-Source Edge Technologies with Odin Chen of Arm 14:49

12 weeks ago14:49

14:49

What if you could deploy AI solutions seamlessly across every platform with just a single stroke? Join us as we explore this intriguing possibility alongside our guest, Odin Chen of Arm, who brings a wealth of experience from his journey from Taiwan to Cambridge. Together, we embark on a journey through the vibrant landscape of Edge AI, tackling its transformative impact on both enterprises and startups. We shine a light on the mantra "write once, deploy everywhere," unraveling its significance in a world where AI's reach is rapidly expanding. Along the way, we confront the challenges of AI deployment, from soaring costs to energy demands, and celebrate the critical role of open-source projects. Discover how technologies spanning image classification to generative AI are revolutionizing industries, while open-source contributions, particularly in Arm architecture and Python optimizations, are paving the way for a more collaborative future. In the second segment, we traverse the vast potential of the AI ecosystem, shedding light on AI frameworks like MediaPipe and ESSEC torch that guide industries toward optimized AI solutions. With Google's AI offerings on the horizon, we introduce a cutting-edge context model for ESSEC torch and delve into key partnerships with Arm technology pioneers such as Alif and HiMAS. These collaborations aim to turbocharge scalable computing solutions and high-efficiency image processing. Excitingly, we unveil our developer program and open software stack, inviting listeners to engage actively and collaborate. Keep your calendars marked for upcoming live demos and panel discussions, promising a firsthand look at the immense potential of AI deployment across diverse sectors. Send us a text Support the show Learn more about the EDGE AI FOUNDATION - edgeaifoundation.org…

1
Revolutionizing Resource-Constrained Devices with Cutting-Edge HIMAX Edge AI Processors 16:41

13 weeks ago16:41

16:41

Unlock the fascinating world of edge AI technology with insights from Karan Kapoor of Himax Technologies. Discover how Karan and his team are pioneering the development of ultra-low-power AI processors designed for resource-constrained devices. We'll explore their latest innovation, the WE2 processor, which optimizes the balance between model size, speed, energy efficiency, and accuracy. Karan shares the intricacies of integrating hardware and software systems, along with groundbreaking AI applications like facial recognition and keyword spotting, all while staying true to Hymax's mission of making AI cost-effective and accessible for all. In the second segment, we spotlight a transformative partnership that brings Wwiseit2 into Edge Impulse's AI ecosystem. This collaboration is a game-changer, introducing a no-code platform that simplifies data collection and synthetic data generation. We unravel the process from model design to testing, supported by an AutoML feature that ensures precision. With ARM's Ethos U55 powering the AI processor, the episode explores how partnerships with innovators like GrowVision AI are crafting an advanced ecosystem. Developers, stay ahead of the curve by following these cutting-edge developments on platforms like LinkedIn and GitHub. Send us a text Support the show Learn more about the EDGE AI FOUNDATION - edgeaifoundation.org…

1
Renesas and Reality AI: Shaping the Future of Edge Computing 13:15

14 weeks ago13:15

13:15

Discover the cutting-edge world of TinyML and Edge AI with our special guest, Eldar Sido from Renesas. As a key figure in the AI Center of Excellence, Eldar unveils the fascinating challenges and breakthroughs in embedding AI into compact devices like microcontrollers and sensors that are rapidly spreading across the globe. This episode promises an exploration into the dynamic realm of TinyML, focusing on its three foundational pillars: real-time analytics, voice, and vision. Eldar discusses the anticipated double-digit growth of this sector over the next decade, making it a pivotal listen for tech enthusiasts and those curious about the future trajectory of AI technology. Eldar also shares how Renesas, through its acquisition of Reality AI, is enhancing its prowess in real-time analytics and tackling high-frequency sampling rates. With a focus on consultancy, data selection, and sensor placement, Eldar illustrates how only a sliver of the TinyML process is devoted to model development. By integrating Reality AI with the E-Square Studio IDE, Renesas is streamlining workflows to maximize the potential of TinyML. Tune in to understand how TinyML is reshaping industries and what lies ahead for Edge AI applications. This episode is packed with insights and strategies for anyone looking to leverage the power of TinyML in their projects. Send us a text Support the show Learn more about the EDGE AI FOUNDATION - edgeaifoundation.org…

EDGE AI POD

1
Revolutionizing Weather Forecasting with Acoustic Smart Technology 25:04

29 weeks ago25:04

25:04

Unlock the secrets to revolutionary weather forecasting with our latest episode featuring Jonah Beysens, the brilliant mind behind the Aurora project. Imagine a world where remote and resource-limited areas can access reliable weather data without the hassle of maintaining traditional mechanical stations. Learn how the Aurora, an acoustic smart weather station, is making this possible. Born from the TinyML Challenge 2022, Aurora employs a TinyML board with a microphone to classify wind and rain intensity, ensuring robustness and ease of deployment. Jonah walks us through the journey of crafting a real-world dataset and shares the collaborative spirit behind making this data open for community-driven innovation. Discover how the shift towards low-cost, acoustics-based devices is reshaping weather forecasting, offering enhanced spatial resolution with multiple ground stations. Jonah sheds light on the collaborative efforts to refine prediction models with open datasets, emphasizing the profound global impact, especially in developing nations where agriculture depends heavily on accurate forecasts. As we discuss the ongoing work to make weather data more accessible worldwide, we highlight the role of community and open access in driving forward weather-related technologies. Join us in exploring how these innovative solutions promise timely, dependable forecasts, paving the way for a future where environmental data is a shared resource for all. Send us a text Support the show Learn more about the EDGE AI FOUNDATION - edgeaifoundation.org…

EDGE AI POD

1
Revolutionizing Nano-UAV Technology with Cutting-Edge On-Device Learning Strategies 21:57

30 weeks ago21:57

21:57

Unlock the secrets of edge computing and on-device learning with Elias Cereda, a trailblazing PhD student at the Dalle Molle Institute for Artificial Intelligence in Italy, as we explore the transformative potential of AI in nano-UAV technology. Discover how Elias is pioneering solutions to the challenges of domain shifts in AI models, particularly those trained in simulations and deployed in real-world scenarios. From enhancing privacy and security through on-device learning to mastering the intricacies of human pose estimation for autonomous navigation, this episode is a treasure trove of insights. We dive into the practicalities of deploying systems using the Crazyflie 2.1 drone alongside GreenWaves' sophisticated RISC-V GAP8 and GAP9 chips, all while managing tight power constraints. Take a closer look at the complexities of on-device and online learning with Elias as we delve into the nuances of continual learning. Learn how maintaining a balance between retaining previous knowledge and integrating new data is crucial in the field. Elias introduces us to a self-supervised metric that operates without initial labels, cleverly utilizing odometry data to refine model predictions. By considering human movement as measurement noise and refining data through post-processing techniques, the episode unveils strategies to enhance model performance without explicit tracking. For those eager to delve deeper into the methodologies discussed, Elias points to a comprehensive paper that promises to enrich your understanding of this cutting-edge technology. Send us a text Support the show Learn more about the EDGE AI FOUNDATION - edgeaifoundation.org…

EDGE AI POD

1
Deploying TinyML Models at Scale: Insights on Monitoring and Automation with Alessandro Grande of Edge Impulse 20:34

31 weeks ago20:34

20:34

Unlock the secrets of deploying TinyML models in real-world scenarios with Alessandro Grande, Head of Product at Edge Impulse. Curious about how TinyML has evolved since its early days? Alessandro takes us through a journey from his initial demos at Arm to the sophisticated, scalable deployments we see today. Learn why continuous model monitoring is not just important but essential for the reliability and functionality of machine learning applications, especially in large-scale IoT deployments. Alessandro shares actionable insights on how to maintain a continuous lifecycle for ML models to handle unpredictable changes and ensure sustained success. Delve into the intricacies of health-related use cases with a spotlight on the HIFE AI cough monitoring system. Discover best practices for data collection and preparation, including identifying outliers and leveraging Generative AI like ChatGPT 4.0 for efficient data labeling. We also emphasize the importance of building scalable infrastructure for automated ML development. Learn how continuous integration and continuous deployment (CI/CD) pipelines can enhance the lifecycle management of ML models, ensuring security and scalability from day one. This episode is a treasure trove of practical advice for anyone tackling the challenges of deploying ML models in diverse environments. Send us a text Support the show Learn more about the EDGE AI FOUNDATION - edgeaifoundation.org…

EDGE AI POD

1
Revolutionizing Smart Glasses with Ultra-Efficient AI Algorithms 30:37

32 weeks ago30:37

30:37

Unlock the secrets of ultra-efficient AI with Julian Musman from ETH Zurich, as we explore the cutting-edge world of smart glasses. Discover how Julian and his team have mastered the art of deploying powerful AI algorithms on the tiniest of devices, tackling the monumental challenge of performing object detection on microcontrollers. Tainissimo YOLO, a breakthrough network, is not just a marvel of engineering but a testament to the balance between computational power, network size, and battery life, revolutionizing the way smart glasses operate without relying on the cloud. Join us as we unpack the hardware innovations behind these intelligent glasses, featuring components like the NRF53 microcontroller and the GAP9 neural engine, meticulously designed for energy efficiency and high-speed performance. With an astonishing end-to-end latency of 18 frames per second, these glasses promise all-day functionality on a single charge. Julian sheds light on the intricate harmony of PCB design and neural network architecture and explores the potential of advanced convolutional changes, all while emphasizing the strategic choice to remain loyal to traditional networks for unmatched performance with hardware accelerators. Get ready for an insightful journey into the future of AI-powered wearable technology! Send us a text Support the show Learn more about the EDGE AI FOUNDATION - edgeaifoundation.org…

EDGE AI POD

1
Cloud to the Edge: Future of LLMs w/ Mahesh Yadav of Google 59:48

33 weeks ago59:48

59:48

Curious about how you can run a colossal 405 billion parameter model on a device with a mere 2 billion footprint? Join us with Mahesh Yadav from Google, as he shares his journey from developing small devices to working with massive language models. Mahesh reveals the groundbreaking possibilities of operating large models on minimal hardware, making internet-free, edge AI a reality even on devices as small as a smartwatch. This eye-opening discussion is packed with insights into the future of AI and edge computing that you don't want to miss. Explore the strategic shifts by tech giants in the language model arena with Mahesh and our hosts. We dissect Microsoft's investment in OpenAI’s Phi model and Google's development of Gamma, exploring how increasing the parameters in large language models leads to emergent behaviors like logical reasoning and translation. Delving into the technical and financial implications of these advancements, we also address privacy concerns and the critical need for cost-effective model optimization in enterprise environments handling sensitive data. Advancements in edge AI training take center stage as Mahesh unpacks the latest techniques for model size reduction. Learn about synthetic data generation and the use of quantization, pruning, and distillation to shrink models without losing accuracy. Mahesh also highlights practical applications of small language models in enterprise settings, from contract management to sentiment analysis, and discusses the challenges of deploying these models on edge devices. Tune in to discover cutting-edge strategies for model compression and adaptation, and how startups are leveraging base models with specialized adapters to revolutionize the AI landscape. Send us a text Support the show Learn more about the EDGE AI FOUNDATION - edgeaifoundation.org…

EDGE AI POD

1
Revolutionizing Space Technology with Edge AI and Satellite Autonomy 1:01:39

40 weeks ago1:01:39

1:01:39

Imagine a world where satellites can make real-time decisions without waiting for ground-based commands. Learn how edge AI is revolutionizing space technology with David Maloney from Ubotica, who brings unparalleled expertise from his time at Movidius and Intel. Uncover the transformative power of performing AI inference directly on satellites, and why traditional ground-based systems are no longer sufficient. We explore advanced imaging techniques like hyperspectral imaging, which can now detect forest fires or monitor environmental changes with unprecedented speed and accuracy. Delve into the fascinating intersection of image compression and generative AI, and see how these technologies can overcome the challenges of power consumption and energy efficiency in space. Hear about real-world implementations such as the FISAT-1 satellite's cloud detection and removal capabilities and the impressive radiation tolerance of hardware tested at CERN's Large Hadron Collider. By examining the trade-offs and legacies of space technology, we reveal how cutting-edge AI applications are optimizing data transmission and processing. Finally, we discuss the advanced applications of hyperspectral sensors and image pattern recognition in environmental monitoring and security. Learn how these technologies can assess vegetation health, evaluate soil nutrients, and even detect illegal activities. Discover the challenges of deploying sensors in the harsh conditions of space and the significance of early warning systems for space weather events. With insights into the European Space Agency's Vigil mission and real-time edge processing in space, we also cover the exciting potential of Tiny Train technology for on-board training and updates, ensuring that our satellites are smarter and more responsive than ever before. Send us a text Support the show Learn more about the EDGE AI FOUNDATION - edgeaifoundation.org…

EDGE AI POD

1
Is it time to take control of your Edge AI Project? A conversation with SensiML 56:07

43 weeks ago56:07

56:07

Ready to unlock the secrets behind real-time sensing algorithms for embedded devices? In this episode, you’ll gain valuable insights into transforming raw sensor data into actionable intelligence. We begin by breaking down the complexities of signal preprocessing, event triggering mechanisms, and feature extraction. Learn how event detection can drastically reduce false positives and computational overhead, setting the stage for robust and efficient real-time systems. Next, we delve into the world of AutoML tailored for embedded devices. Discover the nuances of hyperparameter tuning, cross-fold validation, and model ranking based on key performance metrics like F1 score and accuracy. We also introduce you to Piccolo AI, an open-source game-changer for model building and sensor data management. Plus, we answer listener questions about practical implementation details, making this segment a must-listen for any embedded systems enthusiast. Finally, get hands-on with Piccolo AI as we guide you through setting up your environment using Docker and exploring its powerful web interface. From adding new feature extractors to using synthetic data for model validation, we cover everything you need to contribute to or leverage this open-source project. We also emphasize community engagement, highlighting how collaborative efforts can drive innovation and improve the versatility of machine learning in embedded systems. Join us and become part of a thriving community pushing the boundaries of what's possible. Send us a text Support the show Learn more about the EDGE AI FOUNDATION - edgeaifoundation.org…

EDGE AI POD

1
Optimization Techniques for Powerful yet Tiny Machine Learning Models 59:37

49 weeks ago59:37

59:37

Can machine learning models be both powerful and tiny? Join us in this episode of TinyML Talks, where we uncover groundbreaking techniques for making machine learning more efficient through high-level synthesis. We sit down with Russell Clayne, Technical Director at Siemens EDA, who guides us through the intricate process of pruning convolutional and deep neural networks. Discover how post-training quantization and quantization-aware training can trim down models without sacrificing performance, making them perfect for custom hardware accelerators like FPGAs and ASICs. From there, we dive into a practical case study involving an MNIST-based network. Russell demonstrates how sensitivity analysis, network pruning, and quantization can significantly reduce neural network size while maintaining accuracy. Learn why fixed-point arithmetic is superior to floating-point in custom hardware, and how leading research from MIT and industry advancements are revolutionizing automated network optimization and model compression. You'll gain insights into how these techniques are not just theoretical but are being applied in real-world scenarios to save area and energy consumption. Finally, explore the collaborative efforts between Siemens, Columbia University, and Global Foundries in a wake word analysis project. Russell explains how transitioning to hardware accelerators via high-level synthesis (HLS) tools can yield substantial performance improvements and energy savings. Understand the practicalities of using algorithmic C data types and Python-to-RTL tools to optimize ML workflows. Whether it's quantization-aware training, data movement optimization, or the fine details of using HLS libraries, this episode is packed with actionable insights for streamlining your machine learning models. Send us a text Support the show Learn more about the EDGE AI FOUNDATION - edgeaifoundation.org…

EDGE AI POD

1
Revolutionizing Embedded Computer Vision with Cutting-Edge AI Processors 43:50

50 weeks ago43:50

43:50

Unlock the secrets to high-efficiency embedded computer vision with industry experts Sandeep Mistry from Arm and Alex Chang from Himax Technology. We promise insights into the TinyML community and the collaborative efforts that are setting new benchmarks in the field. Alex Chang shares an in-depth look at Himax Technology’s cutting-edge AI processors and always-on sensors designed for low-power, battery-operated devices. Learn how Himax’s dual-mode CMOS sensors are revolutionizing applications like doorbells and smart devices, and why efficient AI processing at the edge is increasingly crucial. Curious about the architecture of state-of-the-art AI processors? Discover the intricate details of a sophisticated Wi-Fi processor designed specifically for AI applications. We explore groundbreaking features such as DC-DC converters, dynamic voltage and frequency scaling (DVFS), and advanced security measures like Physical Unclonable Functions (PUF) and qubit-toe accelerators. You'll also understand how specialized hardware accelerators for motion detection, JPEG decoding, and voice detection enhance comprehensive audio and video AI processing. The development ecosystem, complete with essential software tools and downloadable resources on GitHub, is laid out to ease your implementation process. Ready to see AI in action? Follow our guided process of porting a pose estimation model to the iMac's YSI2 processor, aimed at tasks like fall detection and human-machine interfaces. We detail how to use a modified version of the UltraLytics GLLv8 postmodel for better accuracy and how to optimize the model for the ARM Ethos U55 MPU using the Ethos U Vella compiler. Witness a live demo showcasing the model’s real-time performance, achieving over 10 frames per second. We wrap up with a discussion on revolutionary partnerships driving Edge AI forward, spotlighting contributions from companies like Syntient, Atreus, ARM, and many others. Don't miss this chance to stay at the forefront of embedded computer vision and Edge AI technology! Send us a text Support the show Learn more about the EDGE AI FOUNDATION - edgeaifoundation.org…

EDGE AI POD

1
Pioneering Personalized Medicine with Advanced AI Algorithms 1:06:10

1 year ago1:06:10