Data Center Performance, Edge Efficiency

Axelera Wingman
Your Agentic Partner

Building on new AI hardware usually means writing code, configuring models, and fixing deployment issues one step at a time. Axelera® Wingman changes that. This agentic platform is built on the knowledge base of the Voyager® SDK and turns plain language into working AI. Describe what you want, and Axelera Wingman builds the pipeline, runs it on the hardware, and iterates until it performs.

Contextual Video Intelligence with Vision Language Model

Cameras capture everything, but making sense of it usually takes a human watching, or a model trained for one narrow task. Vision Language Models understand context, answer questions, and summarize scenes in plain language. Ask where a child who wandered off in an amusement park is, and get a real answer, not just a bounding box. Running entirely on Axelera hardware, it keeps video on-site.

High-Throughput Video Analytics, Server-Scale Inference

Analyzing many live camera feeds usually means racks of power-hungry GPUs and heavy cooling. This demo shows another way. A Prodrive server with six Axelera® Server 150p cards runs Ultralytics YOLO26x at 1,200 frames per second on traffic feeds and a live camera. Lower power means a cooler, quieter system, quiet enough to run live right here at the booth.

And there is more to explore. Visit us at booth no. 3251 to see these demos live and talk with our team