AWL Expands Real-Time Edge AI Video Analytics Platform Built on NVIDIA Metropolis
— Turning video into structured data: real-time, on-device analytics that accelerates digital transformation in the field —

AWL, Inc. (Headquarters: Shinagawa-ku, Tokyo; President & CEO: Muneharu Kitade; hereinafter “AWL”), a certified startup originating from Hokkaido University that develops core edge AI technologies in-house and brings them to society as video analytics solutions, today announced that it is strengthening the deployment of its edge AI video analytics solutions built around NVIDIA Metropolis, a platform for vision AI applications.

■ Background

As AI technologies continue to advance and see broader real-world adoption, the reach of AI is extending beyond the digital realm into the physical spaces where people and goods move and interact. Physical AI—AI that perceives and understands the real world and informs decisions and actions on the ground—is drawing global attention as the next major arena for AI. In Japan, meanwhile, labor shortages driven by a declining and aging population continue to intensify, and sites in retail, manufacturing, and other industries face the pressing challenge of maintaining and improving safety, quality, and customer experience with limited staff.
Few data sources capture the state of physical spaces as richly as video from the cameras already installed on-site. While this footage represents a valuable data asset for addressing these challenges, cloud-dependent analytics has faced constraints in terms of latency, network load, and privacy protection. Against this backdrop, edge AI video analytics—which processes video instantly on-site, at the edge, and extracts only the necessary information as structured data—is increasingly regarded as a foundational technology for bringing physical AI into real-world use.

■ Solution Overview: AWLBOX and NVIDIA Metropolis

AWLBOX, AWL's edge AI device, employs edge computing technology based on the NVIDIA Jetson platform and enables AI-powered video analytics simply by connecting to existing cameras. Because it requires no dedicated cameras or large-scale wiring work, it keeps deployment burdens low and is being adopted across a wide range of environments, including retail stores, commercial facilities, and manufacturing sites.
The platform leverages the NVIDIA Metropolis software stack, combining a highly efficient video processing pipeline built on NVIDIA DeepStream with AI model training and optimization using NVIDIA TAO to deliver high-accuracy video analytics. It further incorporates Multi-Camera Multi-Target Tracking (MCMT), which follows the same subject across multiple cameras, and vision language models (VLMs), which interpret video content at a semantic level, to extract structured data describing who is doing what, when, and where.
These processes are optimized with NVIDIA CUDA and TensorRT and run entirely on-device in real time, without sending video to the cloud. This enables low-latency, power-efficient operation, and because video data never leaves the site, the design is also effective from a privacy standpoint—supporting rapid decision-making in the field. AWL will start exploring the further acceleration of their development process with NVIDIA’s new portfolio of agent-ready skills.

■ Applications and Track Record

AWL's solutions have been adopted primarily by retailers in Japan for retail marketing applications such as shopper behavior analysis and the visualization of product interactions, with a proven track record at companies including Satudora. More recently, deployments have been expanding into manufacturing—covering worker flow analysis and enhanced safety management on factory floors—as well as into smart city applications, such as people-flow analytics for revitalizing urban areas and enhancing community safety and security.

■ Outlook

AWL will continue to leverage advanced AI technologies, including NVIDIA Metropolis, to advance the real-time visualization of people and physical spaces—maximizing the value of on-site data and helping customers transform their operations.

■ About AWL, Inc.

AWL is a certified startup originating from Hokkaido University that develops core edge AI technologies in-house and has brought them into real-world use as video analytics solutions. With more than 16,000 edge AI cameras deployed across Japan, primarily in the retail sector, AWL's strength lies not only in research and development but in its proven ability to implement and scale technology in the field. Headquartered in Hokkaido—a region where social challenges such as population aging and labor shortages emerge ahead of the rest of Japan—AWL brings together a global engineering team spanning Japan, Vietnam, and India to deliver high-quality, cost-effective solutions optimized for real-world sites. Its proprietary AI technology complements the human eye, visualizing people and physical spaces in real time, as AWL continues to take on social challenges together with customers worldwide.

Established: June 1, 2016
Representative: Muneharu Kitade, President & CEO
Offices:
[Tokyo HQ] THE CROSS GOTANDA 8F, 2-24-4 Nishi-Gotanda, Shinagawa-ku, Tokyo
[Sapporo HQ] 4-1-20 Kita 8-jo Higashi, Higashi-ku, Sapporo, Hokkaido
Business: Development and provision of edge AI video analytics solutions
URL: https://awl.co.jp