Archives: Livestream

Join the EDGE AI FOUNDATION and our Neuromorphic Working Group for a special livestream exploring how neuromorphic computing, event-based sensing, spiking neural networks, and sparsity-aware architectures are enabling a new generation of efficient and intelligent Edge AI systems.

Bringing together experts from BrainChip, fortiss, Politecnico di Torino, Sony Advanced Visual Sensing, Fraunhofer, CSEM, and DTU, the program will move from emerging research to practical applications and deployment strategies.

Topics will include conditional sparsity and the future of neuromorphic hardware, adaptive event-based perception, continual and federated learning for industrial robotics, privacy-preserving fall detection using event-based vision and neuromorphic processors, spiking neural networks for real-time control, and new approaches to measuring computational efficiency in sparse and spiking neural networks.

Featured presentations include:

From CNNs to Conditional Sparsity: A Deployment Strategy for the Next Decade of Neuromorphic Hardware — BrainChip

Adaptive Closed-Loop Control of Temporal Integration in Event-Based Embedded Perception — fortiss

Continual and Federated Learning of Neuromorphic Gesture Interfaces for Industrial Welding Robots — Politecnico di Torino

Privacy-Preserving Fall Detection at the Edge Using Sony IMX636 Event-Based Vision Sensor and Intel Loihi 2 Neuromorphic Processor — Sony Advanced Visual Sensing

Spiking Neural Networks for Low-Latency, High-Reliability Feedback Control in Power Conversion Systems: The EdgeAI Approach — Fraunhofer Institute for Integrated Circuits / Technische Hochschule Nürnberg

EFLOP: A Sparsity-Aware Metric for Evaluating Computational Cost in Spiking and Non-Spiking Neural Networks — CSEM

Discover how the neuromorphic ecosystem is moving beyond traditional von Neumann computing toward more adaptive, event-driven, energy-efficient AI at the edge.

TL;DR: Roofline and Arm built end-to-end data-tiling for Arm’s SVE in MLIR and IREE, making AI code generation adapt to the vector width found at runtime and delivering an average 30x speedup over the same stack without data-tiling. Because that infrastructure was built independently of any specific ISA, porting it to RISC-V’s RVV took weeks rather than months, with support for hundreds of AI models from day one and better performance than ExecuTorch on 16 of 22 models on the SpaceMiT X60.

Scalable vector ISAs such as Arm’s SVE and RISC-V’s RVV are replacing fixed-width extensions like Arm NEON. They let silicon vendors implement hardware with vector widths that suit their product, while a single binary still runs across all of them, including hardware generations still on the roadmap. This newly won flexibility now requires software support in the compiler: Vector-length-agnostic code generation for AI models needs to adapt tile shapes and memory layouts effectively to the available vector width instead of assuming fixed-width vector targets. To get a scalable software solution that brings the latest AI models to these different chips in no time, the compiler needs to be built with a reusable and adaptive infrastructure.

This webinar shows how Roofline’s infrastructure-based approach using MLIR and IREE allowed us to migrate the stack from one vector extension to another in mere weeks, while supporting hundreds of AI models from day one. Together with Arm, we built end-to-end data-tiling for Arm’s SVE in MLIR and IREE, where tile sizes are derived from the runtime vector length rather than baked in at compile time. The hard part was not generating SVE instructions, but making the compiler stack itself scalable. Data-tiling turned out to matter more than the vectors themselves, with an average speedup of 30x over non-data-tiled SVE code and gains over both IREE’s and ExecuTorch’s backends for the fixed-width vector extension NEON. We could then reuse most of our infrastructure to port our support to RISC-V’s scalable vector extension RVV. Because the infrastructure and our optimizations are independent of the specific ISA, enabling a second scalable target required only weeks rather than the months we needed to build the infrastructure around Arm SVE. Roofline’s RVV support outperforms ExecuTorch on 16 of 22 models on the RISC-V hardware SpaceMiT X60.

You will leave this webinar knowing how the MLIR and IREE infrastructure lets Roofline build a reusable software stack that can be applied across different implementations of scalable hardware, across different ISAs and also largely across different hardware families.

Welcome to our newest installment of EDGE AI Workshops – providing you with hands on knowledge to build the latest edge AI technologies, taught directly from industry and academic experts – FOR FREE.
———————————————

WORKSHOP: Building a Privacy First Voice Kiosk: Local RAG & LLMs on Arduino UNO Q

START TIME: 8:15AM PST

Hosted by Dr. M. Fekri (Moorcheh) & A. Richetta (Qualcomm-Arduino)

NOTE: Arduino is providing 10 UNO Q boards to builders who attend and interact during the live webinar. Attendees will also receive complimentary access to Moorcheh Edge (up to 10,000 vectors). Together, it’s an ideal hardware and software setup to build Moorcheh information theoretic search engine and RAG on Arduino to build your local AI assistants.

AUDIENCE: Edge AI developers and innovators building offline, privacy preserving AI products without cloud dependencies.

GOALS:
(1) Deploy a fully local, hands free AI assistant managing concurrent workloads within a ~3.6 GB RAM limit.

(2) Use Moorcheh Edge 1 bit embedding compression for ~17 ms semantic search on device.

(3) Orchestrate the offline stack: local LLM (Llama 3.2 1B), Whisper (STT), and Piper (TTS).

(4) Transition to a standalone appliance running continuously on USB-C power without cloud APIs.

PREPARATION: None mandatory. To replicate: Arduino UNO Q (4GB), USB mic/speaker, host PC on the same Wi-Fi, and the provided walkthrough repo.

ACTION: Swap the demo catalog for your own data and run the complete open source pipeline with zero cloud fees.

TIME: 1 hour (webinar + live demo)
————————————————–

WORKSHOP: Training Free Food Understanding with Vision Language Models at the Edge

START TIME: 9:30AM PST

Hosted by D. Pau and A. Spagnolo (STMicroelectronics) and A. Nagendra (Qualcomm).

TARGET AUDIENCE: Edge AI developers, computer vision engineers, embedded developers, and technical decision makers interested in running multimodal AI pipelines locally on constrained hardware.

GOALS: By the end of the session, attendees will understand:
(1) How a training free food understanding pipeline combines image text retrieval and vision language models without task specific fine tuning.

(2) How image text retrieval reduces large ingredient and caption vocabularies to a small candidate set before invoking a more expensive VLM.

(3) How confidence based gates skip unnecessary model calls to reduce latency while preserving prediction quality.

(4) How this architecture supports three concrete tasks:

(4.1) Ingredient recognition – retrieve likely ingredients and refine the final selection with Qwen3-VL-4B-Instruct.

(4.2) Dish caption retrieval – retrieve candidate descriptions and rerank them with Qwen3-VL-Reranker-2B.

(4.3) Calorie estimation – estimate ingredient quantities or portion sizes and combine them with nutritional information.

(5) How the full pipeline was optimized and deployed locally on both NVIDIA Jetson AGX Orin and Qualcomm Dragonwing IQ-9075 EVK.

PREPARATION REQUIRED:
(1) Nothing mandatory to follow along.

(2) To reproduce the demo: an NVIDIA Jetson AGX Orin and a Qualcomm Dragonwing IQ-9075 EVK, plus the project repositories and setup instructions shared with attendees.

DURATION: 1 hour (technical presentation + live on device demo)
————————————————–

WORKSHOP: When Buildings Speak in Codes
Neural Codec & On Device Qwen3 for Vibration Diagnostics

Hosts: N. Kordjukovs & D. Pau (STMicroelectronics), A. Nagendra (Qualcomm)

START TIME: 10:45AM PST

AUDIENCE: Edge AI developers and engineers building private, sensor driven diagnostics without sending raw data off device.

GOALS:
(1) PIPELINE: Learn how the codes_v3 pipeline converts 10 second sensor windows into 250 single token symbols per sensor via a residual VQ codec.

(2) COMPARISON: Compare 1 reference STWIN.box (STEVAL-STWINBX1) against 5 targets using codec symbols and broadband levels.

(3) SCENARIOS: Explore live 6 board USB monitoring (running Qwen3-4B on the IQ-9075 NPU), immutable recorded replay, and operator triage.

(4) PRODUCTION: Keep raw data local, hash verify files, apply strict schema validation, and use fail closed error handling.

PREPARATION REQUIRED: None mandatory. Live replay requires: Qualcomm IQ-9075 EVK, 6 STEVAL-STWINBX1 boards, a USB hub, and the provided repo.

Moving Edge AI from a successful pilot to real world production brings a new set of challenges. From hardware and software integration to performance, scalability, security, lifecycle management, and long term maintainability, what works in a proof of concept does not always translate easily to deployment at scale.

In this EDGE AI Talks session, Wind River (Aptiv) will share their perspectives on the key challenges organizations face when bringing AI to devices at the edge and explore how these barriers can be addressed. The discussion will look at what it takes to move beyond experimentation and build robust, scalable, and production ready Edge AI solutions that can be deployed and managed reliably in the real world.

Join us for practical insights into closing the gap between Edge AI pilots and production at scale.

How is Edge AI transforming the operating room and enabling the next generation of precision medicine?

Join us for this EDGE AI Solutions livestream as we explore how real-time AI is enhancing surgical workflows, supporting clinical decision-making, and improving patient outcomes. From intelligent imaging and computer-assisted interventions to low-latency inference and secure on-device processing, Edge AI is unlocking new possibilities where reliability, privacy, and speed are critical.

Our expert panel will discuss the technologies, challenges, and real-world applications driving AI adoption in healthcare, including the role of Edge AI in precision surgery, clinical environments, and future intelligent healthcare systems.

Speakers:
Micha Pfeiffer, National Center for Tumor Diseases (NCT)
Krish Iyer, Applied Research Office of the CTO, Dell Technologies
Connie Hebert, Chief Nursing Officer & Healthcare Subject Matter Expert, Dell Technologies

Whether you’re developing AI solutions for healthcare, advancing medical devices, or interested in the future of intelligent clinical systems, this session will provide valuable insights into how Edge AI is reshaping the future of surgery and patient care.

Join us for an in-depth EDGE AI Talks session with Hellbender and Ambarella as we explore how robotic vision perception systems evolve from early prototypes to production-ready deployments. Learn how Hellbender leverages Ambarella’s Edge AI platform to develop intelligent, high-performance perception for robotics, covering the challenges of real-world deployment, optimization, and scalable Edge AI solutions. Discover practical insights, development workflows, and lessons learned for bringing robotic vision applications from concept to reality.

Watch three back-to-back hands on workshops with Intel, STMicroelectronics, MiTO and ForestHub to learn the latest techniques and technologies in generative and agentic edge AI.——————————————–

WORKSHOP #1- Building your local voice assistant on compact edge device- Hosted by Danilo Pietro Pau, IEEE and ST Fellow & Ashutosh Kumar, Ph.D., M.B.A, AI Technical Marketing Lead at Intel

This workshop explores how to build a fully local, low-latency conversational AI agent optimized for edge devices. Participants will examine a production-ready pipeline integrating streaming Speech-to-Text, a quantized Small Language Model, and real-time Text-to-Speech, orchestrated via OpenVINO across CPU, GPU, and NPU on Intel® Core™ Ultra Processor powered edge device. We highlight advanced optimizations including neural VAD for fast endpointing, overlapping audio windows for accurate transcription, and token-level output streaming with punctuation-aware chunking to reduce time-to-first-audio. Through hands-on insights, developers will learn how to design responsive, power-efficient voice assistants that run entirely on-device—delivering privacy, reliability, and real-time interaction without cloud dependency.

What you will learn:
– Understanding how to quantize AI models with OpenVINO for optimized performance – be it speech to text, text to speech, or language models
– Explore how local machine learning models can be used to run multi-modal AI workloads for man
– Hands-on lab with open-source models and an application to try the workflow out in Jupyter notebook environment

* First 20 participants will get access to an instance to try it out and experience the compute efficiency of Intel® Core™ Ultra processor powered edge devices. Other participants are welcome to download the workshop and try on their own systems.
——————————

WORKSHOP #2: The Self-Driving Home: TinyAgents Cooperating over Tiny A2A on STM32 = Hosted by Marcus Rueb of ForestHub.ai

TARGET AUDIENCE
Embedded & firmware engineers, edge-AI developers, and technical decision-makers building local, sensor-driven products – anyone interested in running multiple cooperating agents on constrained MCUs/MPUs without the cloud.

GOALS
By the end of the session attendees will understand:
– How several TinyAgents run locally on the STM32N6 (Neural-ART NPU) and STM32MP2, each owning a domain – comfort, energy, safety.
– How those agents coordinate over Tiny A2A, a lightweight agent-to-agent protocol for low-power, bandwidth-constrained devices – instead of static rules/scenes.
– How this plays out across three concrete use cases:
* Comfort – a vision agent detects dirt/spills and dispatches the vacuum robot to the spot.
* Energy – agents detect real presence/usage, steer heating & cooling dynamically, and orchestrate bidirectional EV charging (V2H/V2G).
* Safety – on-device vision detects falls/injury (elderly care) and alerts instantly – no camera frame ever leaves the home.
– How to operate these agents in production: deploy, monitor (AgentOps), update and govern them across a fleet of homes – a full lifecycle, 100% local.

PREPARATION REQUIRED
– Nothing mandatory to follow along.
– To replay the demo: an STM32N6 Discovery kit and/or STM32MP2 board plus our walkthrough repo (link shared ahead of time); a Linux host or WSL environment recommended for the local toolchain.

DURATION
1 hour (technical webinar + live demo)

——————————
WORKSHOP #3 – Grounded Video Understanding on the Edge with Small Language Models – Hosted by Andrea Basso, PhD, MITO Tech Ventures

Deploying vision-language AI on edge devices is no longer just a matter of shrinking models, it requires rethinking how they are structured, optimized, and grounded. In this hands-on workshop, we present a practical approach to building efficient multimodal pipelines using small language models, lightweight vision components, and embedded NPUs.

Participants will learn how to optimize llama.cpp, extend multimodal projection layers for flexible resolutions, quantize models, and deploy YOLO on STM32MP2 hardware. The workshop will present the Narrative Camera, a real-world example of a grounded, multi-stage pipeline that transforms raw video streams into structured, human-readable narratives while minimizing hallucinations.

By the end of the workshop, attendees will have the tools and insights to design by themself an STM32MP2 edge-native AI systems that move beyond raw perception to meaningful, interpretable understanding.

How is edge AI transforming commercial HVAC systems?

Join us on the latest EDGE AI Solutions livestream, co-hosted by Avnet, as Engenuity shares how its groundbreaking platform is bringing real-time intelligence directly to rooftop HVAC units (RTUs).

Discover how edge AI enables predictive maintenance, improved energy efficiency, smarter building operations, and greater sustainability—all while moving beyond traditional building management systems.

Learn how intelligent rooftop HVAC is shaping the future of commercial buildings and creating new opportunities for facility owners, operators, and technology providers.

Featured Speaker: David Rozio, CEO, Engenuity

Getting into Edge AI isn’t just about having the right technical skills anymore. As competition for engineering roles continues to grow and hiring processes evolve, understanding how to navigate the job market has become just as important as the technology itself.

In Volume 4 of the EDGE AI Career Livestream in collaboration with 5V Tech, we’ll go inside the hiring process to explore what recruiters and hiring managers are really looking for, why strong candidates still get overlooked, and how engineers can position themselves more effectively in a competitive market.

Our panel brings together perspectives from hiring, academia, and engineering to share practical advice, real world experiences, and an honest look into what it takes to move from application to offer.

Featuring:
🎙️ Rick Morales – Head of Global Talent Acquisition, Ambiq

🎙️ Dr. Tinoosh Mohsenin – Associate Professor, Johns Hopkins University

🎙️ Sakshi Rathi – Senior Machine Learning Engineer & Tech Lead, Apple

Moderated by
Luke Perrins – Senior Consultant, Edge AI at 5V Tech and Chair of the EDGE AI FOUNDATION Career Working Group

What We’ll Explore
» The Reality of Hiring in Edge AI – What hiring managers actually look for, why applications get filtered out, and common mistakes candidates make.
» AI for Resume & LinkedIn Optimisation – How engineers are using AI tools, LinkedIn, and personal branding to improve visibility and strengthen applications.
» What Separates Successful Candidates – Insights into what helps candidates progress through interviews and stand out in a crowded market.
» Future Skills & Career Growth – How hiring is evolving, emerging skills companies are prioritising, and how engineers can stay competitive.

Whether you’re a student, recent graduate, career changer or engineer looking for your next opportunity, this session is designed to give you a clearer understanding of what actually helps candidates stand out and get hired in today’s Edge AI landscape.

You’ll leave with practical strategies, insider perspectives, and a clearer understanding of how to approach your next career move in Edge AI with confidence!

The THOR x NeuroBench Challenge 2026 focuses on advancing Brain-Computer Interfaces (BCIs) by utilizing the Motor Imagery (MI) paradigm.

Participants are tasked with designing event-driven neuromorphic models to classify EEG signals from the OpenBMI dataset into left- or right-hand movements. The goal is to move this technology out of the clinical lab and into wearable decoders for motorized wheelchairs and robotic prosthetics, offering life-changing potential for individuals with stroke, ALS, or spinal cord injuries.

The awards for this challenge are sponsored by the EDGE AI FOUNDATION. This livestream tutorial provides a comprehensive, step-by-step guide to navigating the competition’s submission process.

The session breaks down the specific formatting requirements for the mandatory preprocessed data and walks through a complete example submission. Attendees will receive practical instructions on how to correctly build, package, and submit their solutions using the NeuroBench Code Harness during the Phase-I submission window.