Axelera AI

🌟 Growing the Axelera AI Team 🌟 Over the last few months, we’ve had the pleasure of welcoming some incredible new talent to Axelera AI. From engineering to product, operations to business development, each new team member brings fresh perspectives, unique expertise, and a shared passion for simplifying and accelerating AI at the edge. 💡 Innovation happens when brilliant minds come together, and our growing team is proof of that. A big, warm welcome to all our new colleagues: 🌕 Edward Kuo 🌕 Sean Conlon 🌕 John B. 🌕 Ashish Ramakant Gujarathi 🌕 Ruxanda Foca 🌕 Govardhana Krishna [Kris] 🌕 Clement Dourval 🌕 Francesco Vidaich 🌕 Koby Soden 🌕 Marta Ostroumoff (FCCA) We’re thrilled to have you on board and can’t wait to see the impact you’ll make.

In this must-read interview with Verve Ventures, Fabrizio Del Maffeo shared a few reflections on the company’s journey building an AI chip designed specifically for edge computing. The conversation covers Axelera’s fast-paced development cycles, approach to hiring, and why Europe needs sovereignty in AI hardware. It also touches on how Metis - originally built for vision tasks - is now being used to run compact generative AI models like LLaMA and Phi-3 Mini. Check it out!

Liam is building something seriously exciting: a real-time pose feedback and training assistant for exercise and weight lifting, powered by vision AI and the Metis platform. In his latest progress report, Liam shared how he’s evolved his prototype with some classic web dev thinking - splitting the app into a React.js front-end and an AI-powered back-end running on Axelera’s Metis hardware with the Voyager SDK. 💬 “I tend to learn best by getting my hands dirty first and reading the manual later… Moving to prototype, I learned the Voyager SDK does a lot of the heavy lifting I was doing manually with OpenCV.” Liam’s backend handles AI inference and serves API endpoints for status, video streams, and keypoint tracking, while the frontend renders the UI on the client browser. Early results: ✅ Solid pose tracking ✅ Smooth SDK integration with YOLO ⚠️ Video feed performance still needs work (next stop: learning GStreamer!) We love seeing creative edge AI projects come to life with Metis, and we’re always happy to support the builders behind them. 📣 Have any tips on optimising GStreamer pipelines or experience with live video feeds on embedded systems? Join us on the Axelera AI community. https://lnkd.in/d4mVcbqK #EdgeAI #VisionAI #AIForGood #VoyagerSDK

🔍 Heterogeneous systems are redefining edge AI Axelera AI is proud to sponsor DVClub Eindhoven 2025, taking place on 23 September at the High Tech Campus Eindhoven. This year’s focus? Design & Verification of Heterogeneous Systems—a topic at the heart of what we do. 💡 Why it matters: At Axelera AI, our Metis AI Platform integrates RISC-V, digital in-memory computing, and high-performance AI acceleration—all within flexible form factors like PCIe and M.2. Events like this are vital to strengthening the ecosystem that supports these innovations. 🎟️ Free to attend in person or online - grab your tickets quick! https://lnkd.in/e936aGMz

💡 Project spotlight: Elderly Guardian Brett Moore is working with an Metis AI accelerator system and he’s diving into real-time fall detection using YOLOv8 pose estimation, all processed locally at the edge. His project, Elderly Guardian, is a heartfelt response to a common concern: how to keep seniors safer when loved ones can’t be there in person. Next up? Integrating voice check-ins and smart-home responses, like notifying caregivers or switching on lights after a fall. It’s a beautiful use of AI and smart tech to improve peace of mind for independent people and their families. If you’ve got any tips on YOLOv8 + Metis or creative home automations, come and join in the conversation as Brett's project progresses: https://lnkd.in/er3pSV_N Let’s build tech that cares. 🧠💙 #EdgeAI #SmartHome #ComputerVision #AIforGood #YOLOv8 #Alexa #FallDetection #ElderlyCare

“Axelera AI will become a cornerstone of European AI sovereignty." That is a bold statement. But in just four years, Axelera has developed and launched its Metis AI accelerator chip, raised more than 200 million euros, and grown to more than 200 people. Axelera’s CEO Fabrizio Del Maffeo is exemplary of a new generation of entrepreneurs in Europe. He thinks big, moves fast, and hires missionaries, not mercenaries, as he says in my interview. Read it here: https://lnkd.in/dcUiyQgV If you want to invest in companies like Axelera, join Verve Ventures' growing investor network of people who understand the relevance of such companies for Europe. If you are interested in learning more about the future of computing or playing a part in it, connect with my esteemed colleagues Emma Schepers and Pepijn Rot.

Deploying AI at the edge doesn’t need to be hard. Whether you’re optimizing inference pipelines, tracking performance across hardware configurations, or integrating advanced vision models with minimal overhead, the Axelera AI Metis platform makes it possible. 💡 From real-time analytics to streamlined pipeline building with YAML and GStreamer, we’ve focused on giving developers the tools to benchmark and deploy with confidence. And all this performance and usability is now available direct to you through RUTRONIK Electronics Worldwide. 👉 Find out more: https://lnkd.in/dbnxvfP4

🎉 Axelera AI now supported in the DeGirum AI Hub! We're excited (and not a little bit impressed) to see that Axelera’s Metis platform and Voyager SDK have been integrated into DeGirum’s AI Hub, a browser-based tool that makes it easy to compile #YOLO models for our Metis platform with no toolchains, and no local setup. Upload your PyTorch checkpoint and get an Axelera-optimised binary. Then you can test it in the browser before deploying it. ✅ Supports: • YOLOv8 & YOLO11 (n, s, m, l, x sizes) • Tasks: detection, classification, segmentation, keypoints, oriented boxes • Custom input resolutions • In-browser testing with integrated C++ postprocessing • Fast compilation turnaround This is especially helpful for developers who want to: • Avoid manual toolchain setup • Rapidly test different YOLO variants for Metis • Validate inference outputs before hardware deployment Shashi Kiran Chilappagari has a great post about how you can get started right now over on the community: https://lnkd.in/dz2rz4NZ

Fantastic to hear that RUTRONIK Electronics Worldwide is now carrying Axelera AI’s Metis accelerator boards 😃 This is great for all the AI fans, devs and engineers: 🟡 Quick procurement and stock transparency: You get real‑time availability, pricing in your currency, plus datasheets and lead‑time info directly on Rutronik. 🟡 Developer support tools: The platform offers technical documentation, mass‑quotation uploads and alternative suggestions – useful for rapid prototyping or volume sourcing . Rutronik is more than just a storefront. Go check it out. And thanks for sharing Davin Moorman!

Across all platforms tested in HotTech Vision And Analysis' study, one factor consistently correlated with high performance and low power per frame on large models like YOLOv8l wasn’t raw compute. It was how well the SDK handled multi-stream input and postprocessing out of the box. Platforms that required manual multi-threading, stream handling, or postprocessing pipeline development saw not just longer dev time, but greater runtime inefficiencies, especially on larger models. 🧠 Why it matters: This suggests that for edge AI use cases involving complex object detection across multiple cameras, SDK design and application scaffolding can impact power and throughput as much as the silicon itself. Not just setup speed, but actual inference performance: https://lnkd.in/dZQ-DmwZ