Multi-model person-of-interest ID + weapon detection, built on the Voyager SDK. The "weapon" was a lightsaber (Count Dooku's hilt). The specs were real though! 3x 4-chip Metis cards, 48 AIPU cores → 2.5 PetaOPS, 1,440+ model inferences/sec across multiple 8K streams at
@axeleraai
-
Computex 2026: Live AI Vision, VLM, Voice, Robotics Demos
By
–
One week to #Computex2026. Find us at booth K0104, TaiNEX 1. 6 live demos: 8K on-device vision at 670fps, VLM video search in natural language, AI robot arm on Metis M.2, offline 3-model voice assistant, and more. Book a 1:1 with the team: https://
eu1.hubs.ly/H0vGNSS0 #EdgeAI -
Project GAIUS: Gesture Control Without the Cloud
By
–
Point at a lamp. It turns on. Point at the TV. It changes the channel. No remote, no voice, no cloud. Project GAIUS by @KatusDavid: multi-camera 3D triangulation, arm raycasting, gesture recognition on Metis M.2. Fully offline, fully open source. Code + docs:
https://eu1.hubs.ly/H0vmv_j0 -
Axelera AI Voyager SDK Outperforms Hyperscaler Datacenter GPUs
By
–
You've never heard of Axelera AI's Voyager? It's the SDK that made the Metis run in less than twelve parsecs. It's outrun hyperscaler GPUs. Not the local inference boxes, mind you. I'm talking about the big datacenter racks now. She's fast enough for you, old man…
-
YOLO Edge Deployment with Ultralytics and Metis Partnership
By
–
Our AI friends over at @Ultralytics has put up a new partner page covering YOLO on Metis, which we really want you to take a deep dive into: https://
eu1.hubs.ly/H0tYFwB0 Essential stuff if you're deploying YOLO at the edge:
• One-command export from Ultralytics straight to .axm
• -
Axelera Metis Demo Jam Showcases Developer Innovation
By
–
Our first Demo Jam was supposed to be a low-effort thing. That is not what happened.
The brief was simple: 1 to 2 minute videos of anything built with Axelera Metis and the Voyager SDK. A cool technique, a feature demo, a work-in-progress. We pitched it as low-commitment on -
Edge AI Microscopy System Tracks Paramecium for 5+ Hours
By
–
5 hours 49 minutes. That's how long one live Paramecium was tracked continuously by an unattended edge AI microscopy rig built by Winkoms Open Microscopy.
— Axelera AI (@AxeleraAI) 17 avril 2026
A Metis M.2 running YOLOv8-seg, steering a motorised XY stage to keep a single microorganism centered in the field of view.… pic.twitter.com/hJOGoGDtB35 hours 49 minutes. That's how long one live Paramecium was tracked continuously by an unattended edge AI microscopy rig built by Winkoms Open Microscopy. A Metis M.2 running YOLOv8-seg, steering a motorised XY stage to keep a single microorganism centered in the field of view.
-
Building AI Inference Platform on RISC-V Architecture
By
–
Real lessons from building AI inference on RISC-V. Koby Soden, Solutions Architect at Axelera AI, is speaking at RISC-V Now! on April 21st in San Jose, CA. His session, "Building an AI Inference Platform on RISC-V," covers what it takes to move from edge vision to generative AI
-
Deploy Ultralytics YOLO Models on Axelera Metis AIPUs
By
–
The recording of our live session with @ultralytics and Innowise is now up. See how to deploy Ultralytics YOLO models on Axelera Metis AIPUs in minutes using a single command. Read the technical blog: https://
eu1.hubs.ly/H0tm6sp0
Watch the session: https://
eu1.hubs.ly/H0tm3BT0 #EdgeAI -
Axelera integrates YOLO deployment on specialized AI hardware
By
–
We're looking forward to joining @ultralytics and Innowise today to share a more efficient way to deploy YOLO models. Our new integration allows for deployment on Axelera AIPUs using only a single command. We will be hosting a live demo and may have discounts available for those