EDGE AI POD

What if the control loop could think ahead and correct itself before errors take hold? We dive into a practical leap for motors, inverters, and energy storage: ultra-low-latency edge AI that predicts error trajectories at startup and intervenes inside the loop in about 100 microseconds. Instead of piling on sensors and pushing raw signals to the cloud, we work directly from existing operational data, chart the most efficient path, and act locally—then pass only meaningful transients upstream for fleet analytics and predictive maintenance.We start by grounding the challenge: linear systems tolerate classic PID, but nonlinear dynamics create overshoot, oscillation, and costly performance tradeoffs. Throwing bigger processors at the problem hits limits on cost, memory, and thermals. The solution mirrors a lesson from the smartphone era—where dynamic voltage and frequency scaling transformed performance-per-watt—by bringing adaptive optimization to the plant itself. Our Ultra-Edge technology extends PID behavior into nonlinear territory, shrinking speed error during torque steps and tightening control, even on modest 32 MHz platforms, with further gains as faster silicon comes online.From factory floors to the power grid, the implications are big. In motor drives, torque transitions smooth out with fewer current spikes. In utilities and data centers, grid-forming converters coordinate with renewables and battery energy storage to deliver synthetic inertia, riding through disturbances and supporting stability rather than tripping offline. By acting in microseconds, converters offer a stabilizing boost, enabling higher renewable penetration and a more credible path to net zero. Meanwhile, microcontroller-level filtering trims a million samples per second down to high-value events so teams get signal without noise.If you care about real-time control, nonlinear systems, and scaling stability with clean energy, this conversation brings clear examples, measured results, and a roadmap for adoption—from pilots and soft IP to demo platforms and a growing model library. Subscribe, share with a teammate who owns drives or converters, and leave a review with your biggest control pain point so we can tackle it next.Send us Fan MailSupport the showLearn more about the EDGE AI FOUNDATION - edgeaifoundation.org

The AI landscape is transforming rapidly, and NXP Semiconductors is at the forefront of bringing these capabilities where they matter most—directly to edge devices. Alberto Alvarez delivers a compelling overview of how NXP is enabling sophisticated generative AI to run locally on microprocessors, without relying on cloud connectivity.Unlike companies focused on massive cloud-based AI training, NXP targets the critical deployment phase, where privacy, security, and efficiency are paramount. Their approach empowers developers to create AI-enhanced solutions for industrial automation, healthcare, automotive systems, and smart environments that keep sensitive data completely local.The presentation unveils the EAQ GenAI flow—a comprehensive software pipeline that allows developers to fine-tune and optimize large language models for specific applications without exposing proprietary data to third-party servers. This pipeline includes automatic speech recognition (ASR) based on the Whisper architecture, LLM reasoning with LLAMA3, retrieval-augmented generation (RAG) for domain-specific knowledge, and natural text-to-speech synthesis—all running efficiently on NXP's hardware.Most impressively, through a partnership with Kinara, NXP demonstrates a fully edge-based multimodal AI implementation running on their iMX810 Plus platform. This system combines an 8-billion parameter language model with computer vision capabilities, allowing it to analyze images, reason about visual content, and respond to questions—all without sending any data to the cloud. The implementation achieves remarkable performance metrics, generating 6.5 tokens per second with response latency as low as 1.5 seconds for follow-up questions about images.From robots with enhanced reasoning capabilities to medical assistants that can analyze diagnostic imagery, the possibilities for this technology are vast and expanding daily. As NXP continues pushing the boundaries of what's possible at the edge, they're laying the groundwork for the next frontier: agentic AI systems that can perceive, reason, and act autonomously across multiple modalities.Ready to build secure, private AI applications that don't compromise on capability? Explore NXP's resources and start creating tomorrow's intelligent edge solutions today.Send us Fan MailSupport the showLearn more about the EDGE AI FOUNDATION - edgeaifoundation.org

What if the obstacle to fast, reliable AI isn’t your dataset or your optimizer—but the silicon under your model? We dig into why performance collapses when architecture and hardware don’t align, and we lay out a clear path to ship models that actually fly on the devices your users own. Starting with the Ferrari-and-hummingbird metaphor, we show how theoretical efficiency—FLOPs, parameters, even TOPS—often fails to predict real-world latency, power, and user experience.We walk through a surprising benchmark: MobileNet V2, small and “efficient,” runs slower than an older ResNet18 on GPUs because depthwise, sequential kernels underutilize parallel hardware. Then we zoom out to hardware selection itself, where NPUs can outperform GPUs despite lower TOPS due to operator support, kernel fusion, and memory behavior. The takeaway is simple: architecture matters only in context, and context means the execution engine, compiler stack, and memory hierarchy that will carry your model in production.From there, we share a four-step framework to become hardware aware: profile on real devices from day one, verify operator compatibility early, automate bottleneck discovery and model selection in CI, and optimize with context using hardware-aware pruning and mixed precision. To show how this works in practice, we unpack our Llama 3.2-1B project on Snapdragon Gen 3, where targeted pruning and precision tuning delivered 31% faster token generation, 25% faster prompt processing, and a 126% faster initialization—all with under 1% accuracy loss.If you build models for the edge, mobile, GPUs, or NPUs, this conversation will help you avoid dead-ends and design for the hardware you actually ship on. Subscribe for more deep dives, share this episode with your team, and leave a review to tell us which hardware you’re targeting next.Send us Fan MailSupport the showLearn more about the EDGE AI FOUNDATION - edgeaifoundation.org

Imagine steering a game with nothing but a blink and a glance. That’s the spark behind our latest build: a noninvasive brain-computer interface that runs entirely on a tiny edge microcontroller, translating eye movements into reliable, real-time commands without a laptop or cloud.We start with the human why. Millions live with neurological conditions that constrain movement but preserve eye control—a narrow channel with huge potential. We compare the promises and trade-offs of invasive BCIs like Neuralink, BrainGate, and Synchron against accessible wearables from Emotiv, Muse, and OpenBCI. The big gap is obvious: people need precise, low-latency control without surgery, high cost, or a desktop tether. Our approach uses electrostatic charge sensing with a glasses-ready electrode layout at the nose bridge and a reference behind the ear, capturing strong ocular signals that are practical for daily wear.From there, we break down the full on-device pipeline. A high-pass filter removes drift, a 50 Hz notch kills power-line noise, and a low-pass smooths the signal so a smaller model can focus on meaningful features. A lightweight Z-score event detector stays always-on and wakes the classifier only when something happens, buffering a 300-sample window at 240 Hz across two channels. The classifier is a tiny 1D CNN—convolution, ReLU, pooling, softmax—clocking about 0.76 ms inference with roughly 18 KB flash and 6 KB RAM. With K-fold cross-validation on nine participants, we see around 90% accuracy for four classes: discard involuntary blinks, map voluntary blinks to “click,” and detect left and right glances.We showcase it with a playful demo: blink to jump over obstacles, glance right to change lanes and collect coins. Beyond the fun, the implications are serious—restoring agency with affordable hardware that works off-grid in real time. We close by outlining what’s next: integrating the sensors into everyday glasses, testing across more users and environments, and adding quick calibration for personalization. If accessible control matters to you—whether for assistive tech, gaming, or new hands-free interfaces—this is a glimpse of what near-future wearables can do.Enjoy the episode? Follow the show, share it with a friend, and leave a quick review to help more listeners discover these conversations.Send us Fan MailSupport the showLearn more about the EDGE AI FOUNDATION - edgeaifoundation.org

Semiconductor counterfeiting has grown into a $200 billion annual problem threatening the integrity of global electronics supply chains. As both chip shortages and sophisticated counterfeiting techniques persist, traditional detection methods fall short—requiring complex setups, hardware modifications, or extensive data labeling.Two machine learning engineers from Analog Devices' advanced R&D team unveil their elegant solution: an unsupervised learning approach that captures the unique "fingerprints" of authentic chips by analyzing power signatures during memory operations. What makes their method revolutionary is its lightweight footprint (under 60KB) and ability to run directly on standard Cortex-M4 microcontrollers at the edge, requiring no cloud connectivity or specialized equipment.The team shares their methodology for creating a robust dataset of 1,000 secure authenticator chips and developing a convolutional autoencoder architecture that achieved 100% accuracy in distinguishing authentic components from close counterparts. Their model learns the normal reconstruction patterns of legitimate chips, then flags anomalies when encountering counterfeits with distinctly different power signatures.Beyond secure authenticators, this approach proves universally applicable to any semiconductor from which analog fingerprints can be collected. Rather than replacing traditional cryptographic methods, it serves as an additional security layer that remains effective even when encryption keys might be compromised through side-channel attacks.Ready to strengthen your supply chain against increasingly sophisticated counterfeits? Discover how this scalable, software-based solution could be integrated with your existing security infrastructure to provide an additional layer of protection for critical semiconductor components.Send us Fan MailSupport the showLearn more about the EDGE AI FOUNDATION - edgeaifoundation.org

Your phone, watch, and even your fridge want real-time intelligence—but power and latency won’t tolerate bloated models or generic compute. We walk through a practical path from Python to custom hardware using high-level synthesis, then invite you to prove it in our Efficient Inferencing Hackathon. With a ready-to-run RISC‑V Rocket Core baseline for MNIST, a full Siemens EDA toolchain, and on-demand training, you’ll learn how to cut latency and power while protecting accuracy through precision mapping, parallelism, and smarter dataflow.We start by mapping the compute landscape—CPUs for flexibility, GPUs for throughput, TPUs/NPUs for tensors, and custom FPGA/ASIC designs for peak power-performance-area. From there, we get tactical: use quantization to right-size bit-widths; apply loop pipelining and unrolling to unlock throughput; partition memories and stream between layers to eliminate round-trips; and iterate quickly with HLS directives instead of rewriting RTL. You’ll see how a baseline inference in the millisecond range can be driven far lower with disciplined co-design, and how Catapult HLS, Questa, and PowerPro provide the feedback loop—latency, area, and power—to make confident trade-offs.Participants receive a virtual machine, C kernels for convolution and dense layers, and a step-by-step path from Keras to synthesizable RTL. The goal is simple and demanding: deliver the fastest MNIST implementation that meets accuracy, area, and energy targets. Along the way, the HLS Academy community offers guidance from experts and peers, and winners will be announced at the Edge AI Foundation event in Taipei, with prizes including a 3D printer, an FPGA board, and Bose earbuds.Ready to turn models into efficient silicon? Join the workshop series, claim your VM via the QR code at hls.academy, and use the promo code with two underscores to unlock full access. If this resonates, subscribe, share with a teammate who ships edge AI, and leave a review to help others find the show.Send us Fan MailSupport the showLearn more about the EDGE AI FOUNDATION - edgeaifoundation.org

What if a neighborhood could think, heal, and feed itself? We sit down with James Ehrlich of Stanford to unpack Village OS, a generative AI platform that designs resilient communities by starting with a simple question: what does the land want? From the urban edge of Riyadh to peri-urban sites worldwide, James shows how geospatial data, climate histories, hydrology, and cultural patterns come together to shape housing, farms, energy, and mobility as one living system.We trace James’s path from early game design and digital effects into the world of eco-villages and permaculture, where taste, health, and connection inspired a research agenda: use technology to serve nature and people. The demo moves from contour maps and fluid dynamics to soil restoration, aquaponics, and agrovoltaics that grow shade crops under solar. Real-time modeling toggles apartments, townhomes, and single-family mixes while projecting costs, returns, and service loads for water, energy, and waste. The punchline is elegant: at the neighborhood scale, waste becomes an asset, powering heat, cooling, and purification while closing loops for food and energy security.Funding and measurement get equal attention. Village OS projects ESG and SDG outcomes and carbon sequestration across decades, offering a transparent view for sovereign wealth funds, pensions, and institutional capital. After groundbreak, the operating layer shifts to edge AI: tinyML sensors and small language models form a digital mycelial network with low latency, low energy, and high autonomy, connected by a thin, privacy-safe cloud channel for cross-site learning. It’s resilience defined by human well-being—lower stress, safer streets, access to fresh food, and spaces for elders and children—backed by systems that can ride out disruption.If you care about sustainable housing, regenerative agriculture, microgrids, and the future of edge AI, this conversation offers a practical, hopeful blueprint. Subscribe, share with a friend who’s into systems thinking, and leave a review with the one feature you’d want in your ideal resilient neighborhood.Send us Fan MailSupport the showLearn more about the EDGE AI FOUNDATION - edgeaifoundation.org

Crowded cafés, clinking plates, and echoey halls make conversations exhausting. We set out to change that by fitting real deep learning into an ear-sized device and proving it can separate speech from noise with almost no delay or battery hit. The result isn’t louder sound; it’s clearer lives and less fatigue.We walk through the full Clara enhancement path: transforming raw mic input into log-mel features, stabilizing for gain shifts, and feeding a 40-layer temporal convolutional recurrent network that predicts a mask to preserve voice and suppress noise. Then we show how a light touch of the original signal brings back space and warmth, avoiding the hollow, underwater audio that turns people off. Along the way, we tackle painful transients—the cutlery and clatter that spike hearing aids—and explain how wide dynamic range compression keeps everything comfortable and intelligible.The heart of the story is edge AI done right. Our SPU001 chip uses unstructured sparsity to skip zero multiplies in hardware, shrinking memory needs and power draw by orders of magnitude. That lets a pruned model with effective 10 MB scale run from just one MB of SRAM while holding algorithmic latency near eight milliseconds and total path time under ten. Metrics back it up: higher scale-invariant signal-to-distortion ratios, better hearing aid speech quality scores, and strong user reports. A rapid partnership with New Sound brought this to market in about three months, and audiologists on a noisy show floor heard the difference immediately.If you care about hearing tech, edge computing, or just making conversations effortless again, this one is for you. Hear how small silicon and smart modeling turn “AI” from a buzzword into a daily benefit. Subscribe for more deep dives on practical edge AI, share with someone who struggles in noisy rooms, and leave a review with your toughest audio environment—we might feature it next.Send us Fan MailSupport the showLearn more about the EDGE AI FOUNDATION - edgeaifoundation.org

A failed 5G rollout in a legendary forest forced us to rethink everything we knew about AI infrastructure. Instead of pushing data to distant servers, we turned wearables, sensors, and tiny controllers into a cooperative network that can sense, decide, and act without the cloud. The result is a hands-on tour of decentralized AI: how to split models across devices, why feature fusion matters more than raw horsepower, and what it takes to make ad hoc networks reliable in the wild.We walk through practical patterns for collaboration at the edge, from complementary sensing in search-and-rescue to pooled compute in crowded venues. You’ll hear how we orchestrate parallel processing on microcontrollers, assign inference to one core and radio handling to another, and compress features to keep bandwidth low. We also dig into continual learning and federated averaging, outlining strategies to adapt models locally while protecting privacy and avoiding catastrophic forgetting. Along the way, we share early results from agriculture and public safety pilots, plus the gritty realities of hardware constraints, scarce datasets, and the challenge of testing at scale.If you’re curious about TinyML, edge AI, and how generative models might run collaboratively across many small devices, this conversation lays out a practical path forward. You’ll come away with a clearer picture of when decentralization beats centralized cloud systems, which protocols survive in noisy environments, and why the future of AI may look less like a monolith and more like a swarm. Subscribe, share this episode with a builder who loves constraints, and leave a review to tell us where you’d deploy a swarm of tiny models next.Send us Fan MailSupport the showLearn more about the EDGE AI FOUNDATION - edgeaifoundation.org

How can artificial intelligence transform electric vehicle performance? Discover the groundbreaking application of neural networks to motor control challenges that even Formula 1 legend Michael Schumacher helped identify.The automotive industry's electrification demands increasingly sophisticated silicon solutions, particularly for traction inverters controlling electric motors. Traditional control systems face a fundamental challenge: they must operate at extraordinary speeds (currently 20kHz, trending toward 100kHz) while managing rapid transitions between states. When drivers make sudden accelerator changes, conventional PID controllers produce energy-wasting overshoots that drain precious battery power.Our research presents a novel approach using neural networks to compensate for these limitations. By generating time-varying correction factors, our AI solution reduces maximum overshoots by up to 70% in demanding scenarios. This innovation represents a critical advancement for electric vehicle efficiency, potentially extending range and improving performance.What makes this application particularly fascinating is the extreme time constraints. While most AI applications process data at relatively leisurely rates (think 30 frames per second for vision systems), motor controllers must complete their calculations within microseconds. Our current implementation achieves 70-microsecond inference times on automotive-grade microcontrollers, with further optimizations planned through hardware acceleration.The collaboration between academic researchers and industry partners (MathWorks and STMicroelectronics) demonstrates the power of combining simulation expertise with real-world deployment capabilities. Using Simulink as the development platform and ST's developer cloud for automatic deployment to physical microcontrollers, we've created a streamlined methodology for applying AI to automotive control systems.Want to dive deeper into the technical details? Check out our published research paper on arXiv and discover how neural networks are transforming the heart of electric vehicle propulsion systems. Share your thoughts on how AI might further revolutionize automotive technology!Send us Fan MailSupport the showLearn more about the EDGE AI FOUNDATION - edgeaifoundation.org

Ever shipped a smart device and worried what happens after it leaves the lab? We dig into the hard parts of edge security—where models live on-device, firmware updates are routine, and attackers treat your fleet as a supply chain—then break them down into moves any team can adopt. From secure boot that blocks untrusted code at power-on to verified boot with discrete secure elements, we show how to anchor trust in hardware so software can prove itself before it runs.We walk through the real risks teams face—model theft, OTA hijacking, plaintext credentials in flash, and silent downgrades—and map them to practices that actually scale across mixed hardware. You’ll hear why encrypting data at rest frustrates drive cloning, how end-to-end encrypted and signed updates prevent tampering, and why automatic rollback turns “bricks” into recoverable hiccups. Updating AI models becomes a strength when you ship small, signed artifacts instead of full images, with logs that satisfy CRA and NIS2 audits while giving operators the visibility they need.We also tackle the build-versus-buy dilemma with clear-eyed math. Building a secure update stack across Qualcomm, NXP, PSoC, and diverse compute modules takes specialists and months; a platform approach spreads cost, speeds delivery, and still lets you own your keys so you can switch later without stranding devices. That key ownership underpins true end-to-end trust: you sign, devices verify, and the infrastructure moves at your pace. If you care about safeguarding IP, maintaining uptime, and earning customer trust, this is your blueprint.If this deep dive helps, follow the show, share it with your hardware and firmware teams, and leave a quick review—what part of your edge stack needs the strongest lock?Send us Fan MailSupport the showLearn more about the EDGE AI FOUNDATION - edgeaifoundation.org

Imagine steering a game or selecting a letter with nothing but a blink or a glance. We set out to make that feel normal, not magical, by building a non-invasive brain–computer interface that runs entirely on a low-power microcontroller and fits into everyday wearables like glasses. No surgery, no cloud dependency—just smart sensing, tight signal processing, and a tiny neural net that turns eye movements into reliable commands.We start with the “why”: millions live with motor impairments yet can still move their eyes, leaving a powerful window for communication and control. From there, we map the BCI landscape—high-precision invasive implants like Neuralink, BrainGate, and Synchron on one side; accessible non-invasive tools like Emotiv, Muse, and OpenBCI on the other—and unpack the trade-offs across accuracy, latency, cost, and ethics. Our approach uses electrostatic charge sensing to read subtle changes around the eyes, with electrodes positioned for comfort and signal quality. A lean pipeline cleans the data with high-pass, notch, and low-pass filters; a Z-score event detector wakes the model only when something meaningful happens.The model is a compact 1D CNN that classifies four classes—discard involuntary blinks, trigger with a voluntary blink, and detect left or right glances—achieving about 90% accuracy on a small multi-participant dataset. Running on an STM32H7, it uses roughly 18 KB flash and 6 KB RAM, with sub-millisecond inference; the overall response is driven by the short data window at 240 Hz, delivering real-time control for basic tasks. We demo blink-to-jump and look-to-steer gameplay to prove responsiveness and highlight how the same system could power communication aids and smart-home control. Looking ahead, we focus on integrating the electrodes into comfortable glasses, adding quick calibration for personal variability, and expanding the command set without sacrificing simplicity.If this mix of accessibility, edge AI, and practical human–machine interaction resonates with you, follow the show, share it with a friend, and leave a review so we can reach more builders and caregivers working on assistive tech. What would you control first with a glance?Send us Fan MailSupport the showLearn more about the EDGE AI FOUNDATION - edgeaifoundation.org

What if transformer-level face recognition could run on a microcontroller without giving up speed or accuracy? We set out to make that real on the STM32N6 by pairing its neural processing unit with a hybrid model that blends convolutional efficiency and attention-like global context. Along the way, we rewired core assumptions about attention, reworked unsupported operators, and delivered a full on-device pipeline that actually feels instant.We start with the hardware edge: ARM Cortex M55, 4 MB of continuous RAM, and an NPU pushing up to 600 GOPS at remarkable power efficiency. That lets us chain models—RetinaFace-style detection with landmarks, alignment for a stable canonical view, MobileNetV2 anti-spoofing to block print and replay attacks, and a final recognizer that outputs a 512‑dimensional embedding. The recognizer is built on EdgeFace, itself based on EdgeNext, chosen for its sweet spot between parameter count and accuracy. It behaves like a transformer where it matters—capturing long-range relationships—yet fits into the tight compute envelope of a microcontroller.The turning point is attention without the dot product. Because the ST toolchain doesn’t support batch matmul, we replaced it with a convolutional self-attention mechanism. Depthwise and pointwise convolutions encode relationships across pixels and channels, a sigmoid stands in for softmax, and element-wise products reconstruct attention’s weighting behavior. This maps cleanly to the NPU, avoids quadratic costs, and preserves the ability to stabilize identities across pose, lighting, and occlusion.Benchmarks show roughly 40 ms per frame end to end—about 25 FPS—plus substantial speedups over STM32H7 and higher accuracy than MobileFaceNet across validation sets. That opens doors for privacy-first access control, frictionless enrollment on-device, and personalized experiences where latency matters and data should never leave the edge. If you’re exploring embedded AI, this walkthrough shows how to align model design with silicon capabilities and deliver results that feel both fast and trustworthy.Enjoy the deep dive? Subscribe, share this episode with a fellow edge AI builder, and leave a quick review to help others find the show.Send us Fan MailSupport the showLearn more about the EDGE AI FOUNDATION - edgeaifoundation.org

A cyclist disappears to the model, not to your eyes—and that mismatch is the heart of safety-critical AI. We open with the “vanishing cyclist” to show how tiny, imperceptible perturbations can flip life-or-death decisions, then walk through a practical path to trust that spans data, verification, and deployment. Along the way, we share real stories from BMW, Airbus, and Madrid Metro to ground the engineering in results, not hype.We break down how to build a resilient pipeline: domain-specific data labeling, realistic synthetic generation for rare and risky scenarios, and tight interoperability across MATLAB, Python, PyTorch, TensorFlow, and ONNX. We dig into explainability beyond classification with D-RISE for object detectors and semantic segmentation, helping you see what the network actually uses to decide. Then we raise the bar with formal verification for robustness—mathematical guarantees within defined perturbation sets—so you aren’t mistaking the absence of found attacks for true safety.Finally, we get practical about the edge. Model compression and projection recover accuracy with fewer parameters, enabling fast, power-efficient deployment to CPUs, GPUs, and FPGAs, backed by code generation for the entire application. We also cover runtime safeguards like out-of-distribution detection to catch smog-on-the-runway moments and escalate safely. Throughout, we connect the work to evolving standards, the EU AI Act, and updated workflows that adapt the V-model for learning systems, so your process and artifacts are ready for audits and certification.If you care about trustworthy AI for cars, planes, rail, and medical devices—and want tools and habits that survive contact with reality—this one’s for you. Listen, subscribe, and leave a review with your biggest trust gap or the safeguard you’d ship first.Send us Fan MailSupport the showLearn more about the EDGE AI FOUNDATION - edgeaifoundation.org

Shipping edge AI shouldn’t feel like a marathon through model zoos, missing ops, and latency ceilings. We lay out a practical path to get from your data and constraints to a hardware-ready model—measured on real boards—without the endless back-and-forth between data science and firmware teams. If you’ve wrestled with quantization loss, unsupported kernels, or picking the “right” NPU, this walkthrough will feel like oxygen.We start by naming the pain: quick demos that collapse under real device limits, foundation models that fail after export, and feedback loops that burn months. From there, we unpack Aptos, our automation engine that turns edge AI into a data in, model out process. The system explores parameterized architecture recipes and neural architecture search, trains promising candidates, and deploys them to a hardware farm packed with evaluation kits. Every candidate returns hard numbers—latency, per-layer timing, memory, on-device accuracy, and power—so tradeoffs are grounded in measurements, not wishful thinking.What makes it fast is the learning layer. As Aptos accumulates results, meta models predict runtime, memory fit, and stable hyperparameter ranges before committing compute. That means less time wasted on dead ends and more time converging on models that satisfy your KPIs, whether you care about sub-5 ms inference on an i.MX 8 Plus, battery life in the field, or non-square inputs that match your camera feed. We also fold in research-backed techniques—pruning, quantization, distillation—so you benefit from the latest without chasing papers.If your team is eyeing a chip migration or evaluating new NPUs, a dropdown swap in Aptos triggers a fresh search tuned to the new hardware, minimizing lock-in and keeping options open. The result is timeline compression: where projects used to take 12–18 months with large teams, we aim to surface strong, deployable candidates in one to two weeks. Subscribe for more deep dives into edge AI deployment, share this episode with your team, and leave a review telling us which device you want to target next.Send us Fan MailSupport the showLearn more about the EDGE AI FOUNDATION - edgeaifoundation.org

What if the fastest path to efficient edge AI isn’t a bigger CPU, but a smarter stream of data? We pull back the curtain on NeuralArt—the flexible, stream‑based accelerator inside the STM32N6—and show how a decade of prototypes led us to rethink how tensors move, how layers are scheduled, and how much work a compiler can save when memory is the real bottleneck. Instead of shuttling activations back and forth, our architecture routes data through specialized units in tightly orchestrated “epochs,” keeping compute hot and bandwidth cool.From there, we tackle the hard limits of standard‑cell designs on practical MCU nodes. Power efficiency stuck around 1–5 TOPS/W and density near 0.1–2 TOPS/mm² pushed us to explore in‑memory computing. We break down digital versus analog IMC—determinism and integration on one side, approximate but highly efficient compute on the other—and share prototype results that hit roughly 40 TOPS/W and about 10 TOPS/mm² at 1 GHz. Along the way, we dig into why half of system power can vanish into data movement and how weight‑stationary strategies change the game.We also get candid about trade‑offs. Embedded phase change memory (PCM) brings remarkable density and multi‑level storage, but demands strict weight‑stationary mapping and drift compensation. No single technology wins every metric, so we lay out a heterogeneous 2D mesh that blends digital IMC, analog IMC, and classical stream units. Our compiler assigns each subgraph to the node that fits its accuracy, throughput, and energy needs, and our NeoSoC research effort moves this vision toward silicon with an upcoming 80‑nm tapeout.If you care about edge inference, memory bandwidth, quantization, and real‑world efficiency beyond spec‑sheet peaks, this conversation is for you. Subscribe, share with a teammate who’s wrestling with on‑device AI, and leave a review with the biggest bottleneck you want us to tackle next.Send us Fan MailSupport the showLearn more about the EDGE AI FOUNDATION - edgeaifoundation.org

Sensors are flooding the edge with data while CPUs juggle denoising, formatting, and inference. We built ADA to flip that script: a Turing-complete neuromorphic processor that computes with time-encoded spikes, slashing power, latency, and memory movement by keeping work inside an event-driven pipeline.We start by unpacking why conventional embedded architectures stall under modern workloads, from pre-processing bottlenecks to compromised security on battery-powered devices. Then we break down neuromorphic fundamentals—how spikes encode information and why sparsity matters—and compare general-purpose frameworks, highlighting the trade-offs that often inflate activity or force manual design. From there, we explain why we chose interval coding and how we solved its biggest flaw. By predicting future spike times, ADA avoids per-tick updates, reducing complexity from linear to logarithmic with precision and mapping neatly to simple add, multiply, and shift hardware.You’ll hear how the architecture comes together: a tiny neuron core that fits in modest FPGAs, standard interfaces like UART and AER for DVS cameras, and our Axon SDK that compiles Python, NumPy, or C algorithms into deployable binaries—no neuron micromanagement required. We demo a three-tap FIR filter built from modular primitives and show ADA acting as a programmable pre-processing element for event vision. On the DVS128 gesture dataset, ADA’s spatial-temporal denoising cut downstream compute by over 50%, keeping the pipeline sparse and fast.Security gets equal attention. We extended the primitive set with modulus arithmetic to support polynomial math central to post-quantum cryptography such as Kyber. The result: 5x better power efficiency and a 2.5x improvement in energy-latency product over MCU baselines, with clear paths to reduce latency further. It points to neuromorphic cryptography that protects implants and IoT sensors without sacrificing battery life.Ready to try it? The Axon SDK is publicly available. Give ADA a spin, share your toughest edge workload, and subscribe for more deep dives into neuromorphic computing. If this sparked ideas, leave a review and pass it to a friend building at the edge.Send us Fan MailSupport the showLearn more about the EDGE AI FOUNDATION - edgeaifoundation.org

What if your printer didn’t just spit out pages, but actually understood them? We walk through a hands-on look at multimodal AI on the edge—how visual-language models read layouts, extract tables, translate content, and reformat documents right where data lives, without shipping sensitive files to the cloud. It’s a practical tour from passive peripherals to active intelligence, with real workflows and measurable speedups.We share the architecture behind on-device document intelligence: pre-processing that stabilizes inputs, VLMs that localize and reason over text and images, and post-processing that converts outputs into CSVs, charts, and accessibility-friendly layouts. You’ll hear how Qwen 2.5-VL handles complex visual inputs while maintaining strong language performance, and how a Flux-based diffusion setup enables creative generation and targeted edits—from updating dates in greeting cards to changing borders and colors by prompt. Along the way, we unpack quantization with GGUF to run 7B-class models in tight memory, diffusion sampler and scheduler tuning for latency, and NVIDIA-optimized libraries to squeeze more from modest GPUs.Beyond demos, we dig into business and engineering realities: fine-tuning with enterprise data to reduce hallucinations, building guardrails and fallback paths for reliability, and segmenting large documents to manage VRAM. We also discuss why a companion device—AI PC or smartphone—can orchestrate heavy lifting until printer SOCs catch up, keeping data private and workflows responsive. If you care about document AI, privacy by design, or accessibility features like dynamic type and contrast, this conversation makes the path concrete and actionable.Enjoy the deep dive? Subscribe, share with a colleague who lives in PDFs, and leave a review with the one edge use case you want us to test next.Send us Fan MailSupport the showLearn more about the EDGE AI FOUNDATION - edgeaifoundation.org

What if AI felt like a door you could open, not a wall you had to climb? We dig into how Arduino’s approach—accessibility first, power when you need it—turns the edge AI buzz into a concrete path you can follow, whether you’re a student with a starter kit or an engineer shipping to a fleet.We walk through a practical four-step journey: try AI through no-code experiments, understand it with pre-trained models, train by fine-tuning or starting from scratch with your data, and build something real that lives beyond a demo. Along the way, we unpack a core principle we call “abstraction without obfuscation”—removing friction while keeping the logic transparent—so you can inspect, modify, and truly own the systems you create. That design philosophy shapes everything from our open hardware portfolio (TinyML-friendly MCUs up to Linux-capable MPUs) to our integrations with popular AI frameworks and community-driven libraries.You’ll also hear how cloud-native developer tools streamline the messy middle: browser-based workflows, single-device to fleet deployments, secure OTA updates, data collection for predictive insights, and closed-loop model improvement. Plus, we introduce our AI assistant as a coach that explains code, diagnoses bugs, and helps optimize for memory and speed—turning dead ends into learning moments. Real-world validation from a 35-million-strong community and enterprise teams, including automotive innovators, shows how openness and cohesion accelerate the leap from idea to production.If you care about AI that empowers rather than intimidates, this conversation lays out the playbook. Subscribe, share with a teammate who loves to build, and leave a review telling us the project you’re dreaming about—we might feature it next.Send us Fan MailSupport the showLearn more about the EDGE AI FOUNDATION - edgeaifoundation.org

If you’ve ever shipped a model that flew in the cloud and crawled on a device, this conversation is a relief valve. We bring on Andreas from Embedl to unpack why edge AI breaks in the real world—unsupported ops, fragile conversion chains, misleading TOPS—and how to fix the loop with a unified, device-first workflow that gets you from trained model to trustworthy, on-device numbers in minutes.We start with the realities teams face across automotive, drones, and robotics: tight latency budgets on tiny chips, firmware that lags new ops, and the pain of picking hardware without reliable performance data. Instead of guesswork, Andreas demos Embedl Hub, a web platform and Python library that standardizes compilation, static quantization, and benchmarking, then runs your models on real hardware through integrated device clouds. The result is data you can act on: average on-device latency, estimated peak memory, compute-unit usage, and detailed, layer-wise latency charts that reveal bottlenecks and fallbacks at a glance.You’ll hear how to assess quantization safely with PSNR (including layer-level drift), why pruning and optimization must be hardware-aware, and how a consistent pipeline across ONNX/TFLite/vendor runtimes tames today’s fragmented toolchains. We also compare Embeddle Hub’s scope to broader end-to-end platforms, touch on non-phone targets available via Qualcomm’s cloud, and talk roadmap: more devices, deeper analytics, and invitations for hardware partners to plug in.If you care about edge AI benchmarking, hardware-aware optimization, ONNX/TFLite compilation, layer-wise profiling, and choosing devices with data instead of hope, you’ll leave with a practical playbook and a tool you can try today—free during beta. Listen, subscribe, and tell us the next device you want to see in the cloud lab. Your model isn’t done until it runs on real hardware.Send us Fan MailSupport the showLearn more about the EDGE AI FOUNDATION - edgeaifoundation.org

Clothes that quietly listen to your breath might be the missing link between hospital‑grade vigilance and everyday comfort. We walk through how our team built a textile‑integrated breath sensor that actually works in the wild—embroidered interconnects, 3D‑printed dielectric islands, and a carbonized‑silicon yarn strain gauge stitched into a belt—then taught it to estimate breathing at the edge with TinyML.We dig into the engineering choices that matter: why flexible interconnects are the “holy grail” for wearables, how a simple peak detector falls apart with drift and burn‑in, and what it takes to turn raw strain signals into reliable features. After screening public datasets that didn’t match our sensor, we built our own: band‑pass filtering in the 0.1–1 Hz range, three‑second windows, normalization, and event‑button labeling for clean ground truth. From there, we used Edge Impulse’s EON Tuner to search architectures and landed on two contenders—a CNN on time‑domain windows and a compact DNN with wavelet features—then deployed both on an STM32L4 with DMA, timers, and CMSIS‑DSP preprocessing.The results are candid and practical. The CNN was slower but consistently more accurate and robust; the DNN was snappier with lower power but less reliable under offset and noise. Models trained on a different sensor’s data struggled to generalize to our belt, reinforcing a core lesson for smart textiles: sensor‑specific datasets and fine‑tuning are essential. We close by mapping next steps—expanding our dataset, improving transfer across garments and users, exploring hydration prediction, and tightening on‑device optimization—so remote patient monitoring can be seamless, private, and wearable all day.If you enjoy deep dives into edge AI, embedded systems, and human‑centric health tech, follow the show, share it with a colleague, and leave a quick review to help others find it.Send us Fan MailSupport the showLearn more about the EDGE AI FOUNDATION - edgeaifoundation.org

What if the only way to get real gains at the edge is to redesign everything—from the silicon atoms to the app you deploy? That’s the bet Professor-Founder Mohammed Ali made with EMAS, and the results are striking: continuous inference at milliwatts, microsecond wake/sleep cycles, and real benchmarks that hold up against the best in class while burning a fraction of the energy.We walk through how a RISC-V core, dual AI accelerators, and an MRAM/RRAM-backed memory system work together to keep weights on-chip, slash data movement, and power-gate aggressively without losing state. The compiler handles pruning, quantization, and on-the-fly compression to achieve around 1.3 bits per weight without torpedoing accuracy, while a custom memory controller mitigates non-volatile quirks like endurance and read variability. Instead of chasing TOPS, the stack optimizes bandwidth, dataflow, and timing to match the realities of sensors and batteries.The story gets especially interesting with drones. Since propellers—not processors—dominate energy use, EMAS applies tiny AI to the control problem, redistributing load across rotors in real time and extending flight endurance by 60% or more in hardware-in-the-loop simulations. We also dig into wearables and time-series workloads like ECG, audio, and vibration, where sparse sampling pairs perfectly with microsecond power gating. If you build at the edge, the dev experience matters: you’ll hear about the virtual dev kit with remote access to real silicon, a compact evaluation board with modular sensors, and an SDK that plugs into TensorFlow, PyTorch, and Zephyr. Advanced users can map trained models via a CLI; newcomers can lean on a NAS-based flow that proposes architectures meeting strict memory and power budgets.If you care about edge AI, battery life, and shipping reliable products, this conversation is a blueprint for co-designing across the stack to unlock 10–200x energy gains without giving up performance. Subscribe, share with a teammate who owns your edge roadmap, and leave a review with the one use case you’d optimize first.Send us Fan MailSupport the showLearn more about the EDGE AI FOUNDATION - edgeaifoundation.org

What if your model pipeline started with a simple goal—your dataset, your target chip, and your latency or energy budget—and ended with measured results on real hardware? We sit down with Model Cat CEO Evan Petritis to explore how AI can build on-device AI through a closed loop that’s grounded in silicon, not estimates or hopeful benchmarks. From a live demo to a tour of their “chip farm,” we dig into how the platform searches architectures, tunes hyperparameters, and validates performance using vendor kernels and compilers across MCUs, MPUs, and specialized accelerators.We share the story behind the rebrand from Eta Compute to Model Cat and why the shift matters: AI research moves too fast for traditional, component-by-component toolchains. Evan breaks down five pillars for trustworthy, autonomous model creation—closed-loop goals, reality grounding, system-level intent, modular learning from new research, and a single-step, transparent experience. You’ll hear how teams can upload datasets, get automated analytics on splits and distribution shifts, set constraints like sub–5 ms inference or energy per inference, and see success predictions before training even starts.The demo highlights the silicon library and how each device is profiled in depth—supported ops, kernel speeds, memory footprints—so accuracy, latency, and energy are measured on the actual target. Results come as clear Pareto trade-offs with downloadable artifacts that reproduce on-device. We also field audience questions on exporting to Keras and TFLite, supporting time-series and audio keyword spotting, integrating labeling partners, onboarding new MCUs and accelerators, and the roadmap toward neuromorphic targets and cost estimation.If you care about edge AI, embedded ML, and shipping models that meet real-world constraints, this conversation shows a practical path forward: use AI to navigate the fire hose of research, then prove it on silicon. Enjoy the episode—and if it sparks ideas, subscribe, leave a review, and share it with a teammate who lives in notebooks but dreams in devices.Send us Fan MailSupport the showLearn more about the EDGE AI FOUNDATION - edgeaifoundation.org

We share new data showing why drivers see generative AI as a defining force in mobility and how edge inference makes cars faster, safer, and more personal. We map the use cases, hardware shifts, and the move to software-first procurement with clear guidance for builders.• survey highlights on generative AI as a mobility megatrend• definitions and examples of circular economy in vehicles• priority edge use cases in ADAS, safety, and infotainment• hidden value in predictive maintenance and intrusion detection• why inference runs on the edge for latency and reliability• constraints around cost, memory, and over-the-air updates• NPU rise over GPU and evolving CPU roles• software-first buying and model portability trade-offs• smarter sensors, radar AI, and neuromorphic paths• hybrid architectures for sensor fusion and efficiencySend us Fan MailSupport the showLearn more about the EDGE AI FOUNDATION - edgeaifoundation.org

Tired of choosing between performance and freedom? We sit down with Stefan Crossin, CEO and co‑founder of YASP, to unpack how a hardware‑aware AI compiler can speed up training, simplify deployment, and finally make model portability real. The story starts with a distributed team in Freiburg and Montreal and moves straight into the heart of the problem: most AI groups burn time on infrastructure and juggle separate stacks for training and inference, all while staying tethered to one dominant vendor’s software ecosystem.Stefan lays out a different path. YASP converts models into a clean intermediate representation, plugs into the tools teams already use, and applies a closed‑loop optimization system that learns the target hardware. Instead of forcing a new language or workflow, a few lines of integration unlock dynamic kernel generation, graph‑level tuning, and one‑click deployment to different chips, clouds, or edge devices. The result is a practical bridge between “write once” ideals and real‑world performance, where being hardware‑aware—not hardware‑bound—delivers speed without lock‑in.We also dive into the market dynamics behind portability. Incumbents protect moats; challengers need bridges. Cloud providers fear shorter runtimes but win when customers get more value per dollar and per watt. With credible benchmarks showing meaningful gains in training and inference, YASP is courting chip makers, CSPs, and end users through a focused beta, a clear roadmap to launch, and a business model that combines free access with subscription tiers. If you’ve been waiting for proof that AI can be both faster and freer across architectures, this conversation makes the case with clarity and detail.Enjoy the episode? Follow the show, share it with a colleague, and leave a quick review—what platform or accelerator would you target first with true portability?Send us Fan MailSupport the showLearn more about the EDGE AI FOUNDATION - edgeaifoundation.org

What if the smartest part of AI isn’t in the cloud at all—but right next to the sensor where data is born? We pull back the curtain on the rapid rise of edge AI and explain why speed, privacy, and resilience are pushing intelligence onto devices themselves. From self‑driving safety and zero‑lag user experiences to battery‑friendly wearables, we map the forces reshaping how AI is built, deployed, and trusted.We start with the hard constraints: latency that breaks real‑time systems, the explosion of data at the edge, and the ethical costs of giant data centers—energy, water, and noise. Then we dive into the hardware leap that makes on‑device inference possible: neural processing units delivering 10–100x efficiency per watt. You’ll hear how a hybrid model emerges, where the cloud handles heavy training and oversight while tiny, optimized models make instant decisions on sensors, cameras, and controllers. Using our BLERP framework—bandwidth, latency, economics, reliability, privacy—we give a clear rubric for deciding when edge AI wins.From there, we walk through the full edge workflow: on‑device pre‑processing and redaction, cloud training with MLOps, aggressive model optimization via quantization and pruning, and robust field inference with confidence thresholds and human‑in‑the‑loop fallbacks. We spotlight the technologies driving the next wave: small language models enabling generative capability on constrained chips, agentic edge systems that act autonomously in warehouses and factories, and neuromorphic, event‑driven designs ideal for always‑on sensing. We also unpack orchestration at scale with Kubernetes variants and the compilers that unlock cross‑chip portability.Across manufacturing, mobility, retail, agriculture, and the public sector, we connect real use cases to BLERP, showing how organizations cut bandwidth, reduce costs, protect privacy, and operate reliably offline. With 2026 flagged as a major inflection point for mainstream edge‑enabled devices and billions of chipsets on the horizon, the opportunity is massive—and so are the security stakes. Join us to understand where AI will live next, how it will run, and what it will take to secure a planet of intelligent endpoints. If this deep dive sparked ideas, subscribe, share with a colleague, and leave a review to help others find the show.Send us Fan MailSupport the showLearn more about the EDGE AI FOUNDATION - edgeaifoundation.org

Ready to revolutionize your approach to edge AI? Keith Kressin, a veteran with 13 years at Qualcomm before joining MemoryX, shares a breakthrough technology that's transforming how AI operates in resource-constrained environments.MemoryX has developed an architecture that defies conventional wisdom about AI acceleration. Unlike traditional systems dependent on memory buses and controllers, their solution features autonomous parallel cores with localized memory, eliminating bottlenecks and enabling linear scaling from small devices to powerful edge servers. The result? About 20 times better performance per watt than common alternatives like NVIDIA's Jetson platform, all packaged in a simple M.2 form factor that consumes just half a watt to two watts depending on workload.What truly sets MemoryX apart is their software approach. While many AI accelerators require extensive model optimization, MemoryX offers one-click compilation for over 4,000 models without modifications. This accessibility has opened doors across industries – from manufacturing defect detection to construction safety monitoring, medical devices to multi-camera surveillance systems. The technology proves particularly valuable for "brownfield" computing environments where legacy hardware needs AI capabilities without complete system redesigns.The company embodies efficiency at every level. While competitors have raised $250+ million in funding, MemoryX has built their complete hardware and software stack with just $60 million. This resourcefulness extends to their community approach – they offer free software, extensive documentation, and support educational initiatives including robotics camps and hackathons.Curious about bringing AI acceleration to your next project? Visit MemoryX's developer hub for free resources and examples, or purchase their M.2 accelerator directly through Amazon. Whether you're upgrading decades-old industrial equipment or designing cutting-edge multi-camera systems, this plug-and-play solution might be exactly what you need.Send us Fan MailSupport the showLearn more about the EDGE AI FOUNDATION - edgeaifoundation.org

Most conversations about edge computing gloss over the enormous challenge of actually deploying and managing software on constrained devices in the field. As Jason Shepherd, Atym's founder, puts it: "I've seen so many architecture diagrams with data lakes and cloud hubs, and then this tiny little box at the bottom labeled 'sensors and gateways' - which means you've never actually done this in the real world, because that stuff is some of the hardest part."Atym tackles this challenge head-on by bringing cloud principles to devices that traditionally could only run firmware. Their revolutionary approach uses WebAssembly to enable containerization on devices with as little as 256 kilobytes of memory - creating solutions thousands of times lighter than Docker containers.Founded in 2023, Atym represents the natural evolution of edge computing. While previous solutions focused on extending cloud capabilities to Linux-based edge servers and gateways, Atym crosses what they call "the Linux barrier" to bring containerization to microcontroller-based devices. This fundamentally changes how embedded systems can be developed and maintained.The impact extends beyond technical elegance. By enabling containers on constrained devices, Adam bridges the skills gap between embedded engineers who understand hardware and firmware, and application developers who work with higher-level languages and AI. A machine learning engineer can now deploy models to microcontrollers without learning embedded C, while the embedded team maintains the core device functionality.This capability becomes increasingly crucial as edge AI proliferates and cybersecurity regulations tighten. Devices that once performed simple functions now need to run sophisticated intelligence that may come from third parties and require frequent updates - a scenario traditional firmware development approaches cannot efficiently support.Ready to revolutionize how you manage your edge devices? Explore how Atym's lightweight containerization could transform your edge deployment strategy.Send us Fan MailSupport the showLearn more about the EDGE AI FOUNDATION - edgeaifoundation.org

The landscape of artificial intelligence is experiencing a profound transformation, with AI capabilities moving from distant cloud servers directly to edge devices where your data lives. This pivotal shift isn't just about running small models locally—it represents a fundamental reimagining of how we interact with AI systems.In this fascinating exploration, Dell Technologies' Aruna Kolluru takes us deep into the world of edge-deployed AI agents that can perceive their surroundings, generate language, plan actions, remember context, and use tools—all without requiring cloud connectivity. These aren't simple classification systems but fully autonomous digital partners capable of making complex decisions where your data is generated.Discover how miniaturized foundation models like Mistral and TinyLlama, combined with agentic frameworks and edge-native runtimes, have made this revolution possible. Through compelling real-world examples, Aruna demonstrates how these systems are transforming industries today: autonomous factory agents detecting defects and triggering interventions, rural healthcare assistants providing offline medical guidance, disaster response drones generating situational awareness, and personalized retail advisors creating real-time offers for shoppers.The technical journey doesn't stop at deployment. We examine the sophisticated optimization techniques making these models edge-friendly, the memory systems enabling contextual awareness, and the planning frameworks orchestrating multi-step workflows. Importantly, we tackle the critical governance considerations for these autonomous systems, including encrypted storage, tool access control, and comprehensive audit logging.Whether you're a developer looking to build edge AI solutions, an enterprise decision-maker exploring AI deployment options, or simply curious about where AI is headed, this episode offers invaluable insights into a technology that's bringing intelligence directly to where it's needed most. Subscribe to our podcast and join the conversation about the future of AI at the edge!Send us Fan MailSupport the showLearn more about the EDGE AI FOUNDATION - edgeaifoundation.org

When personal tragedy strikes, some find a way to transform pain into purpose. Such is the remarkable story behind Ambient Scientific, where founder GP Singh's mission to prevent falls after losing a family member evolved into groundbreaking semiconductor technology enabling AI at the ultra-low power edge.The journey wasn't simple. Creating chips that could run sophisticated deep learning algorithms on tiny batteries proved more challenging than building data center processors. This demanded innovation at every level – from custom instruction sets and compilers to complete software stacks. What emerged wasn't just a single-purpose chip but a programmable platform with the versatility to support diverse applications while consuming a fraction of the power of conventional solutions.Most fascinating is what GP calls the "gravitational pull" toward edge computing. Applications initially deployed in the cloud inevitably migrate closer to where data originates – from data centers to on-premises, to desktops, to mobile devices, and ultimately to tiny wearables. This migration stems from fundamental business concerns: operating costs, data sovereignty, vendor lock-in, and the inherent distrust organizations have for cloud dependencies. The evidence? In hundreds of customer conversations, GP has yet to meet a single organization content with keeping their AI exclusively in the cloud.Ready to explore ultra-low power AI? Ambient Scientific offers development kits accessible to anyone familiar with embedded systems programming and Python-based deep learning. Join the revolution bringing intelligence to where data is created, not where it's processed. Your next innovation might be powered by a chip that sips power while delivering remarkable AI capabilities.Send us Fan MailSupport the showLearn more about the EDGE AI FOUNDATION - edgeaifoundation.org

Power efficiency has become the new currency in AI, and no company exemplifies this shift better than EMAS. Founded by Professor Mohamed Ali as a spinoff from his groundbreaking research at NTU Singapore, this innovative startup is revolutionizing edge AI with semiconductor technology that delivers unprecedented power efficiency for battery-constrained devices.The story begins in 2018 when Ali and his team set out to examine the entire computing stack from applications down to nanotechnology devices. Their research led to a remarkable breakthrough: a chip architecture that brings memory and compute components closer together, resulting in power efficiency 10-100 times better than competing solutions. Unlike other processors that claim low power consumption only during standby, EMAS's chip maintains ultra-low power usage while actively processing data—the true measure of efficiency for AI applications.Mark Gornson, CEO of EMAS's Semiconductor Division, brings 46 years of industry experience to the team, having worked with giants like Intel and ON Semiconductor. After seeing the benchmarks of EMAS's technology, he came out of retirement to help commercialize what he recognized as a game-changing innovation perfectly timed for the edge AI explosion.The applications are vast and growing. Drones can achieve dramatically longer flight times with lighter batteries. Wearable devices gain extended battery life without compromising functionality. Agricultural equipment benefits from real-time monitoring without frequent recharging. Industrial machinery can be equipped with predictive maintenance capabilities that identify subtle anomalies in vibration, temperature, or current draw before failures occur. Robotics systems gain critical safety features through microsecond decision-making capabilities.For developers, EMAS has prioritized accessibility by ensuring compatibility with familiar frameworks like TensorFlow and PyTorch. Their backend engine handles the translation to optimized binaries, eliminating the learning curve typically associated with specialized hardware.Ready to experience this breakthrough technology? EMAS offers development kits for hands-on testing and even provides remote access to their hardware for preliminary evaluation. See them in person at upcoming industry events in Amsterdam and Taipei, where they'll showcase how their innovative approach is redefining what's possible with battery-powered intelligent devices.Join the edge AI revolution and discover how EMAS is making efficient intelligence accessible everywhere it matters.Send us Fan MailSupport the showLearn more about the EDGE AI FOUNDATION - edgeaifoundation.org

"Do you trust your AI models? Honestly, I don't trust them. We should not trust them." These powerful words from STMicroelectronics' Mounia Kharbouche perfectly capture the security challenge facing the edge AI world today.As organizations rush to deploy AI workloads at the edge, a complex security landscape emerges that demands careful navigation. This fascinating panel discussion dives deep into the three major threat vectors organizations must prepare for: algorithmic attacks that manipulate model behavior, physical attacks on hardware, and side-channel analysis that can steal proprietary models in mere hours.Through vivid examples—like specially designed glasses that can fool facial recognition systems—the panelists demonstrate how seemingly minor vulnerabilities can lead to major security breaches. They explore the security paradox of edge deployment: while distributing AI provides resilience against single points of failure, it simultaneously creates numerous potential attack surfaces requiring protection.The conversation reveals a critical tension between economics and security that often drives deployment decisions. Organizations frequently prioritize cost considerations over comprehensive security measures, sometimes with devastating consequences. All panelists emphasize that security must be a fundamental consideration from the beginning of any AI project, not an afterthought tacked on at deployment.Looking to the future, the discussion turns to emerging threats like agentic AI, where autonomous agents might access resources without proper security constraints. The panel concludes with a sobering examination of post-quantum cryptography and why organizations must prepare now for threats that may not materialize for years but will target systems deployed today.Whether you're developing edge AI solutions or implementing them in your organization, this discussion provides essential insights for securing your systems against current and future threats. Join us to discover how to balance innovation with protection in the rapidly evolving world of edge AI.Send us Fan MailSupport the showLearn more about the EDGE AI FOUNDATION - edgeaifoundation.org

The humble printer - that device gathering dust in the corner of your office - is about to undergo a remarkable transformation. Thanks to advancements in generative AI, printers and scanners are evolving from passive endpoints into intelligent document processing powerhouses.Arniban from Wipro Limited unveils how visual language models (VLMs) like QN 2.5 VL and LayoutLMv3 are being deployed directly on edge devices rather than in the cloud. This breakthrough approach addresses critical data privacy concerns while eliminating the need for continuous network connectivity - perfect for sensitive enterprise environments where document security is paramount.These multimodal AI implementations enable remarkable capabilities that were previously impossible. Imagine a printer that can automatically extract complex tables from documents and convert them into visually appealing charts. Or one that can intelligently correct errors, translate content between languages, adapt layouts for visually impaired users, or even remove advertisements when printing web pages - all without sending your data to external servers.The technical implementation involves clever optimizations to run these sophisticated models on relatively constrained hardware. Through techniques like 4-bit quantization, image downscaling, and leveraging NVIDIA's optimized libraries, these models can function effectively on devices with 16GB of GPU memory - bringing AI intelligence directly to the point where documents are produced.While challenges remain in handling large documents and managing the thermal constraints of embedded devices, this technology marks the beginning of a new era in intelligent document processing. The days of printers as "dumb" input-output machines are numbered. The future belongs to intelligent endpoints that understand what they're printing and can transform it in ways that add tremendous value to users.Try imagining what your workflow could look like when your printer becomes your intelligent document assistant. The possibilities are just beginning to unfold.Send us Fan MailSupport the showLearn more about the EDGE AI FOUNDATION - edgeaifoundation.org

The technological revolution sparked by generative AI threatens to create the deepest digital divide we've ever seen. In this illuminating talk, Danilo Pau from STMicroelectronics reveals how only a handful of companies worldwide possess the resources to fully harness large-scale generative AI, while the rest of humanity risks being left behind.Pau takes us through the sobering reality of today's AI landscape: hyperparameterized models requiring nuclear power plants for training, hundreds of millions in costs, and worrying environmental impacts. But rather than accept this centralized future, he presents a compelling alternative path – bringing generative AI to edge devices.Through a comprehensive survey of recent research, Pau demonstrates that generative AI is already running on edge devices ranging from smartphones to microcontrollers. His team's work with STMicroelectronics processors showcases practical implementations including style transfer, language models, and perhaps most impressively, an intelligent thermostat capable of natural language interaction with reasoning capabilities.What emerges is a vision for AI not as another backend classifier but as a transformative interface between humans and machines. "GenAI is not for another detector," Pau explains. "We need to offer new added value" through natural interactions that understand context and can reason about the world.For researchers and developers, this talk provides concrete pathways to explore: from audio processing as a "low-hanging fruit" to visual question answering systems that run on minimal hardware. The future of AI isn't just in massive data centers – it's in the devices all around us, waiting to be unleashed through energy-efficient processing and innovative approaches to model optimization.Ready to join the movement bringing AI capabilities to everyone? Explore how edge-based generative AI could transform your products and help bridge the growing digital divide.Send us Fan MailSupport the showLearn more about the EDGE AI FOUNDATION - edgeaifoundation.org

What happens when trillions of AI agents can discover, communicate, and collaborate across organizational boundaries? Pradyumna Shari from MIT Media Lab unveils NANDA (Networked AI Agents in a Decentralized Architecture), a groundbreaking open protocol that could fundamentally transform how we interact with artificial intelligence.Drawing a fascinating parallel between computing history and our AI trajectory, Pradyumna explains how we've evolved from isolated large language models to action-capable agents that can reason and act in the world. Yet despite this progress, we're still missing the crucial infrastructure that would allow these agents to find and collaborate with each other across organizational boundaries – essentially, an "Internet of AI Agents."Using a relatable birthday party planning scenario, Pradyumna demonstrates how interconnected agents could effortlessly coordinate calendars, groceries, and bakery orders without human micromanagement. But enabling this vision requires solving complex challenges around agent discovery, authentication, verifiability, and privacy that differ significantly from traditional web architecture.At the heart of NANDA's approach is a three-layer registry system designed specifically for dynamic, peer-to-peer agent interactions. The demonstration showcases how this architecture enables diverse communications – from personal agents that adapt messages between family members to commercial interactions between customers and businesses, all while supporting different communication protocols like Google's A2A and Anthropic's MCP.What makes NANDA particularly exciting is its commitment to democratic, open-source development. Rather than dictating standards, the project invites collaboration from academic and industry partners to build this agent ecosystem together, ensuring it remains transparent, trustworthy, and accessible to all.Visit nanda.mit.edu to learn more about how you can contribute to this vision of a decentralized, collaborative future for artificial intelligence.Send us Fan MailSupport the showLearn more about the EDGE AI FOUNDATION - edgeaifoundation.org

Stepping into a future where AI doesn't require the cloud, NXP is revolutionizing edge computing by bringing generative AI directly to microprocessors. Alberto Alvarez offers an illuminating journey through NXP's approach to private, secure, and efficient AI inference that operates entirely at the edge.The heart of NXP's innovation is their EAQ GenAI Flow, a comprehensive software pipeline designed for iMX SoCs that enables both fine-tuning and optimization of AI models. This dual capability allows developers to adapt openly available Large Language Models for specific use cases without compromising data privacy, while also tackling the challenge of memory footprint through quantization techniques that maintain model accuracy. The conversational AI implementation creates a seamless experience by combining wake word detection, speech recognition, language processing with retrieval-augmented generation, and natural speech synthesis—all accelerated by NXP's Neutron NPU.Most striking is NXP's partnership with Kinara, which introduces truly groundbreaking multimodal AI capabilities running entirely at the edge. Their demonstration of the LAVA model—combining LLAMA3's 8 billion parameters with CLIP vision encoding—showcases the ability to process both images and language queries without any cloud connectivity. Imagine industrial systems analyzing visual scenes, detecting subtle anomalies like water spills, and providing spoken reports—all while keeping sensitive data completely private. With quantization reducing these massive models to manageable 4-bit and 8-bit precision, NXP is making previously impossible edge AI applications practical reality.Ready to experience the future of edge intelligence? Explore NXP's application code hub to start building with EIQ GenAI resources on compatible hardware and discover how your next project can harness the power of generative AI without surrendering privacy or security to the cloud.Send us Fan MailSupport the showLearn more about the EDGE AI FOUNDATION - edgeaifoundation.org

The boundary between humans and computers is blurring as AI capabilities advance, creating opportunities for more natural, conversational interactions with our devices. Raymond Lowe from Intel takes us on a journey through the evolution of human-computer interaction, from simple mouse clicks to sophisticated chatbots that understand context, process images, and engage in meaningful dialogue.At the heart of this transformation is OpenVINO, Intel's toolkit for optimizing neural networks across diverse hardware. Raymond demonstrates how this technology enables edge devices—from laptops to specialized processors—to run sophisticated AI models locally without requiring cloud connectivity. The examples are compelling: generating beautiful images of teddy bears in just seconds on standard laptop GPUs, running large language models that once consumed 25GB of RAM on modest hardware, and creating smart cameras that can describe what your baby is doing without complex coding.Memory management emerges as the hero of this story. Through techniques like quantization (reducing model precision from 32-bit to 8-bit or even 4-bit), OpenVINO dramatically shrinks model size while maintaining accuracy. This isn't just about fitting models into limited memory—it's about activating specialized hardware instructions that can deliver 2-3x performance improvements, transforming sluggish experiences into fluid, real-time interactions.The impact extends beyond technical achievements. Raymond shares the emotional moment when he first got a chatbot running locally: "I never felt so alive when I saw the machine talking to me." For developers, this means being able to create prototypes in weeks rather than months, accessing hundreds of pre-optimized examples, and focusing on building experiences rather than struggling with technical hurdles.Through partnerships with Microsoft's AI Foundry program, these capabilities are being integrated directly into Windows, ensuring consumers get optimal AI performance from their hardware without additional setup. For industries embracing AI—from healthcare to retail to smart cities—OpenVINO offers a path to enhance existing applications while exploring new possibilities at the intersection of traditional and generative AI approaches.Want to experience this revolution yourself? Check out Intel's extensive library of notebooks and examples, or try the Open Edge Platform to start building immediately. The future of human-computer interaction isn't just coming—it's already here on your local device.Send us Fan MailSupport the showLearn more about the EDGE AI FOUNDATION - edgeaifoundation.org

The growing gap between rapidly evolving AI models and lagging deployment frameworks creates a significant challenge for edge AI developers. Maurice Sersiff, CEO and co-founder of Germany-based Roofline AI, presents a compelling solution to this problem through innovative compiler technology designed to make edge AI deployment simple and efficient.At the heart of Roofline's approach is a retargetable AI compiler that acts as the bridge between any AI model and diverse hardware targets. Their SDK supports all major frameworks (PyTorch, TensorFlow, ONNX) and model architectures from traditional CNNs to cutting-edge LLMs. The compiler generates optimized code specifically tailored to the target hardware, whether it's multi-core ARM systems, embedded GPUs, or specialized NPUs.What truly sets Roofline apart is their unwavering commitment to comprehensive model coverage. They operate with a "day zero support" philosophy—if a model doesn't work, that's considered a bug to be fixed within 24 hours. This approach enables developers to use the latest models immediately without waiting months for support. Performance benchmarks demonstrate the technology delivers 1-3x faster execution speeds compared to alternatives like Torch Inductor while significantly reducing memory footprint.Maurice provides a fascinating comparison between Roofline's compiler-based approach for running LLMs on edge devices versus the popular library-based solution LLama.cpp. While hand-optimized kernels currently maintain a slight performance edge, Roofline offers vastly superior flexibility and immediate support for new models. Their ongoing optimization work is rapidly closing the performance gap, particularly on ARM platforms.Interested in simplifying your edge AI deployment while maintaining performance? Explore how Roofline AI's Python-integrated SDK can help you bring any model to any chip with minimal friction, enabling true innovation at the edge.Send us Fan MailSupport the showLearn more about the EDGE AI FOUNDATION - edgeaifoundation.org

The future of AI isn't just in massive cloud servers—it's already sitting in your pocket. In this eye-opening presentation, Yeon-seok, CEO and co-founder of JTIC AI, reveals how his company is revolutionizing the AI landscape by tapping into the underutilized Mobile Processing Units (MPUs) that have been standard in smartphones since 2017.While tech giants pour billions into cloud infrastructure, JTIC AI has identified a critical opportunity: leveraging the powerful AI processors already in billions of devices worldwide. This approach delivers not just cost savings, but crucial advantages including offline functionality, enhanced data security, and real-time responsiveness—without depending on internet connectivity.The technical journey involves three essential components: hardware utilization, model optimization, and runtime software. Yeon-seok breaks down sophisticated model optimization techniques like pruning, quantization, and knowledge distillation that make complex AI models deployable to mobile devices. However, the biggest challenge isn't hardware capability but software fragmentation. Unlike the GPU market dominated by NVIDIA and CUDA, mobile devices operate in a fragmented ecosystem where Apple, Qualcomm, MediaTek, and others maintain incompatible software stacks—creating significant barriers for AI engineers.JTIC AI's innovative solution is an end-to-end automated pipeline that handles everything from model optimization to device-specific benchmarking. Their system can determine which runtime will deliver optimal performance for specific models on specific devices—something that's impossible to predict without comprehensive testing. With this approach, developers can deploy sophisticated AI across the mobile ecosystem without wrestling with manufacturer-specific implementations.Ready to unlock the AI capabilities already sitting in your users' pockets? Discover how on-device AI can transform your applications with better privacy, offline functionality, and faster response times—all while reducing your cloud infrastructure costs.Send us Fan MailSupport the showLearn more about the EDGE AI FOUNDATION - edgeaifoundation.org

Memory bottlenecks, not computational limitations, are the true barrier holding back Edge AI. This revelation lies at the heart of Anaflash's revolutionary approach to intelligent edge computing – a breakthrough AI microcontroller with embedded flash memory that transforms how we think about power efficiency and cost in smart devices.The team has engineered a solution that addresses the two fundamental challenges facing Edge AI adoption: power efficiency and cost. Their microcontroller features zero-standby, power-weight memory with 4-bit per-cell embedded flash technology seamlessly integrated with computation resources. Unlike traditional non-volatile memory options that demand extra processing steps and offer limited storage density, this technology requires no additional masks and scales efficiently.At the core of this innovation is the Near Memory Computing Unit (NMCU), which establishes a tight coupling with flash memory through a wide I/O interface on a single chip. This architecture eliminates the need to fetch data from external memory after booting or waking from deep sleep – a game-changing feature for battery-powered devices. The NMCU's sophisticated three-part design enhances parallel computations while minimizing CPU intervention: control logic manages weight addresses and buffer flow, 16 processing elements share weights through high-bandwidth connections, and a quantization block efficiently converts computational results.Fabricated using Samsung Foundry's 28nm standard logic process in a compact 4 by 4.5 mm² die, the microcontroller delivers impressive results. Testing with MNIST and Deep Auto Encoder models demonstrates accuracy levels virtually identical to software baselines – over 95% and 0.878 AUC respectively. The overstress-free waterline driver circuit extends flash cell margins, further enhancing reliability and performance.Ready to transform your Edge AI applications with technology that combines unprecedented efficiency, performance, and cost-effectiveness? Experience the future of intelligent edge computing with Anaflash's embedded flash microcontroller – where memory and computation unite to power the next generation of smart devices.Send us Fan MailSupport the showLearn more about the EDGE AI FOUNDATION - edgeaifoundation.org

Ever wondered how the electric vehicles of tomorrow will squeeze every last drop of efficiency from their batteries? The answer lies at the fascinating intersection of artificial intelligence and motor control.The electrification revolution in automotive technology demands increasingly sophisticated control systems for permanent magnet synchronous motors - the beating heart of electric vehicle propulsion. These systems operate at mind-boggling speeds, with control loops closing every 50 microseconds (that's 20,000 times per second!), and future systems pushing toward 10 microseconds. Traditional PID controllers, while effective under steady conditions, struggle with rapid transitions, creating energy-wasting overshoots that drain precious battery life.Our groundbreaking research presents a neural network approach that drastically reduces these inefficiencies. By generating time-varying compensation factors, our AI solution cuts maximum overshoots by up to 70% in challenging test scenarios. The methodology combines MatWorks' development tools with ST's microcontroller technology in a deployable package requiring just 1,700 parameters - orders of magnitude smaller than typical deep learning models.While we've made significant progress, challenges remain. Current deployment achieves 70-microsecond inference times on automotive-grade microcontrollers, still shy of our ultimate 10-microsecond target. Hardware acceleration represents the next frontier, along with exploring higher-level models and improved training methodologies. This research opens exciting possibilities for squeezing maximum efficiency from electric vehicles, turning previously wasted energy into extended range and performance. Curious about the technical details? Our complete paper is available on arXiv - scan the QR code to dive deeper into the future of smart motor control.Send us Fan MailSupport the showLearn more about the EDGE AI FOUNDATION - edgeaifoundation.org

The gap between science fiction and reality is closing rapidly. Remember when talking to computers was just a fantasy in movies? Raymond Lo's presentation on building chatbots with OpenVINO reveals how Intel is transforming ordinary PCs into extraordinary AI companions.Imagine generating a photorealistic teddy bear image in just eight seconds on your laptop's integrated GPU. Or having a natural conversation with a locally-running chatbot that doesn't need cloud connectivity. These scenarios aren't futuristic dreams – they're happening right now thanks to breakthroughs in optimizing AI models for consumer hardware.The key breakthrough isn't just raw computational power but intelligent optimization. When Raymond's team first attempted to run large language models locally, they didn't face computational bottlenecks – they hit memory walls. Models simply wouldn't fit in available RAM. Through sophisticated compression techniques like quantization, they've reduced memory requirements by 75% while maintaining remarkable accuracy. The Neural Network Compression Framework (NNCF) now allows developers to experiment with different compression techniques to find the perfect balance between size and performance.What makes this particularly exciting is the deep integration with Windows and other platforms. Microsoft's AI Foundry now incorporates OpenVINO technology, meaning when you purchase a new PC, it comes ready to deliver optimized AI experiences out of the box. This represents a fundamental shift in how we think about computing – from tools we command with keyboards and mice to companions we converse with naturally.For developers, OpenVINO offers a treasure trove of resources – hundreds of notebooks with examples ranging from computer vision to generative AI. This dramatically accelerates development cycles, turning what used to take months into weeks. As Raymond revealed, even complex demos can be created in just two weeks using these tools.Ready to transform your PC into an AI powerhouse? Explore OpenVINO today and join the revolution in human-computer interaction. Your next conversation partner might be sitting on your desk already.Send us Fan MailSupport the showLearn more about the EDGE AI FOUNDATION - edgeaifoundation.org

Logic beats arithmetic in the machine learning revolution happening at Newcastle University. From a forgotten Soviet mathematician's work in the 1960s to modern embedded systems, Settle Machine represents a paradigm shift in how we approach artificial intelligence.Unlike traditional neural networks that rely on complex mathematical operations, Settle Machine harnesses Boolean logic - simple yes/no questions similar to how humans naturally think. This "white box" approach creates interpretable models using only AND gates, OR gates, and NOT gates without any multiplication operations. The result? Machine learning that's not only understandable but dramatically more efficient.The technical magic happens through a process called Booleanization, converting input data into binary questions that feed learning automata. These finite state machines work in parallel, creating logical patterns that combine to make decisions. What's remarkable is the natural sparsity of the resulting models - for complex tasks like image recognition, more than 99% of potential features are automatically excluded. By further optimizing this sparsity and removing "weak includes," Newcastle's team has achieved astonishing efficiency improvements.The numbers don't lie: 10x faster inference time than Binarized Neural Networks, dramatically lower memory footprint, and energy efficiency improvements around 20x on embedded platforms. Their latest microchip implementation consumes just 8 nanojoules per frame for MNIST character recognition - likely the lowest energy consumption ever published for this benchmark. For edge computing and IoT applications where power constraints are critical, this breakthrough opens new possibilities.Beyond efficiency, Settle Machine addresses the growing demand for explainable AI. As regulations tighten around automated decision-making, the clear logical propositions generated by this approach provide transparency that black-box neural networks simply can't match. Ready to explore this revolutionary approach? Visit settlemachine.org or search for the unified GitHub repository to get started with open-source implementations.Send us Fan MailSupport the showLearn more about the EDGE AI FOUNDATION - edgeaifoundation.org

The AI revolution isn't just for tech giants with unlimited computing resources. Small and medium enterprises represent a crucial frontier for edge generative AI adoption, but they face unique challenges when implementing these technologies. This fascinating exploration takes us into an unexpected application: smart laboratory mouse cages enhanced with generative AI.Laboratory mice represent valuable assets in pharmaceutical research, with their welfare being a top priority. While fixed-function AI already monitors basic conditions like water and food availability through camera systems, the next evolution requires predicting animal behavior and intentions. By analyzing just 16 frames of VGA-resolution video, this edge-based system can predict a mouse's next actions, potentially protecting animals from harm when human intervention isn't immediately possible due to clean-room protocols.The technical journey demonstrates how generative AI can be scaled appropriately for edge devices. Starting with a 240-million parameter model (far smaller than headline-grabbing LLMs), the team optimized to 170 million parameters while actually improving accuracy. Running on a Raspberry Pi 5 without hardware acceleration, the system achieves inference times under 300 milliseconds – and could potentially reach real-time performance (30ms) with specialized hardware. The pipeline combines three generative neural networks: a video-to-my model, an OPT transformer, and a text-to-speech component for natural interaction.This case study provides valuable insights for anyone looking to implement edge generative AI in resource-constrained environments. While currently limited to monitoring single mice, the approach demonstrates that meaningful AI applications don't require supercomputers or billion-parameter models – opening doors for businesses of all sizes to harness generative AI's potential.Send us Fan MailSupport the showLearn more about the EDGE AI FOUNDATION - edgeaifoundation.org

Sony's revolutionary IMX500 stands at the forefront of a quiet revolution in edge computing and smart city technology. This isn't just another image sensor—it's the first to integrate AI processing directly on the chip, transforming how visual data becomes actionable intelligence while preserving privacy and minimizing infrastructure requirements.The power of this innovation lies in its elegant simplicity. Rather than sending complete images to cloud servers or external GPUs for processing, the IMX500 performs AI inference locally and transmits only the resulting metadata. This approach slashes bandwidth requirements to mere kilobytes, dramatically reduces power consumption, and—perhaps most critically—protects individual privacy by ensuring that identifiable images never leave the device. For urban environments where surveillance concerns often clash with safety imperatives, this represents a breakthrough compromise.Real-world deployments already demonstrate the technology's transformative potential. In Lakewood, Colorado, where a one-mile stretch of road had become notorious for traffic fatalities, Sony's solution achieved 100% performance in identifying dangerous situations—outperforming three competing technologies while costing less. Through partnership with ITRON, these sensors can be seamlessly deployed using existing streetlight infrastructure, creating mesh networks of intelligent sensors without requiring expensive new installation work or dedicated power sources. This practical approach to deployment makes citywide implementation financially viable even for budget-constrained municipalities.The implications extend far beyond traffic monitoring. From retail analytics to manufacturing quality control, the same core technology can be applied wherever visual intelligence provides value. By bringing AI to the edge in a form factor that addresses privacy, power, and practical deployment challenges, Sony has created a foundation for the next generation of smart infrastructure. Explore how this technology could transform your environment—whether an urban center, commercial space, or industrial facility—by leveraging the power of visual intelligence without the traditional limitations.Send us Fan MailSupport the showLearn more about the EDGE AI FOUNDATION - edgeaifoundation.org

Revolutionizing edge computing just got easier. This eye-opening exploration of ST Microelectronics' ST-IoT Craft platform reveals how everyday developers can now harness the power of artificial general intelligence without writing a single line of code.The modern IoT landscape presents a paradox: billions of devices generate zettabytes of valuable data, yet transforming that raw information into intelligent systems remains frustratingly complex. ST's innovative low-code/no-code platform elegantly solves this problem by distributing intelligence across three key components: smart sensors with embedded AI algorithms, intelligent gateways that filter data transmission, and cloud services that handle model training and adaptation.At the heart of this revolution is truly remarkable in-sensor AI technology. Imagine sensors that don't just collect data but actually think – detecting whether a laptop is on a desk or in a bag, whether an industrial asset is stationary or being handled, or whether a person is walking or running. These decisions happen directly on the sensor itself, dramatically reducing power consumption and network traffic while enabling real-time responses. The platform offers 31 different features including mean, variance, energy in bands, peak-to-peak values, and zero crossing that can be automatically selected and applied to your data.What makes ST-IoT Craft truly accessible is its browser-based interface with six pre-built examples spanning industrial and consumer applications. Users can visualize sensor data in real-time, train models with a single button click, and deploy finished solutions directly to hardware – all without diving into complex code. The platform even handles the intricate details of filter selection, feature extraction, window length optimization, and decision tree generation automatically.Ready to transform your IoT projects with embedded intelligence? Visit stcom, search for ST-IoT Craft, and discover how you can teach your sensors to think – no coding required.Send us Fan MailSupport the showLearn more about the EDGE AI FOUNDATION - edgeaifoundation.org

Protecting valuable AI models from theft is becoming a critical concern as more computation moves to edge devices. This fascinating exploration reveals how sophisticated attackers can extract proprietary neural networks directly from hardware through side-channel attacks - not as theoretical possibilities, but as practical demonstrations on devices from major manufacturers including Nvidia, ARM, NXP, and Google's Coral TPUs.The speakers present a novel approach to safeguarding existing hardware without requiring new chip designs or access to proprietary compilers. By leveraging the inherent randomness in neural network training, they demonstrate how training multiple versions of the same model and unpredictably switching between them during inference can significantly reduce vulnerability to these attacks.Most impressively, they overcome the limitations of edge TPUs by cleverly repurposing ReLU activation functions to emulate conditional logic on hardware that lacks native support for control flow. This allows implementation of security measures on devices that would otherwise be impossible to modify. Their technique achieves approximately 50% reduction in side-channel leakage with minimal impact on model accuracy.The presentation walks through the technical implementation details, showing how layer-wise parameter selection can provide quadratic security improvements compared to whole-model switching approaches. For anyone working with AI deployment on edge devices, this represents a critical advancement in protecting intellectual property and preventing system compromise through model extraction.Try implementing this stochastic training approach on your edge AI systems today to enhance security against physical attacks. Your valuable AI models deserve protection as they move closer to end users and potentially hostile environments.Send us Fan MailSupport the showLearn more about the EDGE AI FOUNDATION - edgeaifoundation.org

Beyond raw computational power lies the true measure of AI system effectiveness. Austin Lyons, founder of ChipStrat and analyst at Creative Strategies, challenges us to rethink how we evaluate Edge AI technologies in this thought-provoking talk on metrics that truly matter.For too long, the industry has obsessed over Trillion Operations Per Second (TOPs) as the gold standard measurement. Lyons expertly deconstructs this limited view, introducing us to a more nuanced framework that considers what users actually experience. As generative AI moves to edge devices, shouldn't we care more about tokens per second—how quickly systems respond to our prompts—than abstract computational capabilities?But speed alone doesn't tell the whole story. What happens when your lightning-fast AI assistant drains your battery in an hour? Lyons presents "tokens per second per watt" as a crucial metric for practical, everyday AI use. He also introduces the concept of "vibes"—those harder-to-quantify qualities like perceived intelligence and personality that make or break user adoption, drawing a compelling parallel to why people choose Apple products despite comparable technical specs from competitors.The most valuable insight comes from Lyons' call for cross-functional collaboration in AI system design. When hardware engineers, software developers, designers, and product managers work in isolation, optimizing for their preferred metrics, the end result often disappoints users. By approaching AI development holistically, teams can make informed trade-offs that deliver better overall experiences—sometimes with less powerful but more efficient models.Ready to transform how you think about AI performance? Subscribe to Austin's newsletter at chipstrat.com where he regularly shares insights on the evolving intersection of semiconductors, AI, and product strategy.Send us Fan MailSupport the showLearn more about the EDGE AI FOUNDATION - edgeaifoundation.org

What if we could put powerful AI anywhere—from underwater cameras monitoring fish to the phone in your pocket? That's the vision driving AI Zip, a company building ultra-efficient AI models that can run on virtually any device.During this fascinating conversation, we explore how AI Zip is pioneering a different path in artificial intelligence by focusing on extreme compression and efficiency. While most companies pursue ever-larger models requiring massive cloud infrastructure, AI Zip is shrinking intelligence to fit where it's needed most—at the edge where 99% of data originates.The numbers are staggering: edge devices collectively possess about 100 times more computing power than all cloud resources combined, yet 95% of AI workloads run in the cloud. This disconnect represents an enormous untapped opportunity that AI Zip is addressing through innovations in model compression and deployment.We dive into real-world applications, including an award-winning smart fish farming solution developed with SoftBank that uses underwater computer vision to optimize feeding and dramatically reduce waste. This practical example shows how specialized AI can deliver enormous value when deployed directly at the data source.Perhaps most thought-provoking is the efficiency gap between artificial and natural intelligence. While our most advanced AI systems require thousands of watts of power, the human brain operates on just 20 watts—about the same as a smartphone. Similarly, a jumping spider can navigate complex 3D environments with millions of neurons, while autonomous vehicles need billions of parameters. Closing this three-orders-of-magnitude efficiency gap represents an exciting frontier for AI research.The future of AI won't just be about bigger models—it will be hierarchical, with specialized intelligence at every level. Subscribe now to hear more conversations with pioneers who are reimagining what's possible at the intersection of AI and edge computing.Send us Fan MailSupport the showLearn more about the EDGE AI FOUNDATION - edgeaifoundation.org

What happens when we redesign computing hardware to work more like the human brain? The results are transformative for edge AI.Sumit Kumar from Inatera takes us inside the world of neuromorphic computing – a revolutionary approach that's bringing brain-like intelligence directly to sensors. Born from research at Delft University of Technology, Inatera is tackling one of the most significant challenges in modern technology: how to perform complex AI tasks on battery-powered devices without draining power.The key lies in spiking neural networks that are fundamentally different from conventional AI approaches. These event-driven networks operate with computational dynamics that mimic brain function, resulting in models 100 times smaller than traditional AI while consuming just a fraction of the power. For applications like video doorbells, acoustic scene classification, and wearable healthcare, this means continuous monitoring with millisecond latency at just a few milliwatts – outperforming traditional microcontrollers by at least 10x.Beyond current applications, neuromorphic computing opens entirely new possibilities. The technology excels not just with conventional vision but with radar sensors and other modalities, particularly in privacy-sensitive situations. Robotics represents another frontier, where neuromorphic systems can enhance environmental perception, process complex sensor fusion, and enable low-latency control. Through academic partnerships and industry collaboration via the Edge AI Foundation, Inatera is helping build the ecosystem that will make neuromorphic computing as ubiquitous as neural networks are today. The future of edge AI may indeed be neuromorphic.Send us Fan MailSupport the showLearn more about the EDGE AI FOUNDATION - edgeaifoundation.org