MulticoreWare, a Californian technology company providing software and engineering services, has collaborated with AMD to develop an AI-powered smart self-checkout for retail.

This innovation enhances retailer security and customer shopping experience through gen AI and computer-vision-based automation. It operates multiple AI models simultaneously, using embedded processor cores such as CPU, GPU and NPU available on AMD Ryzen Embedded 8000 series processors, the first AMD embedded devices to integrate NPUs alongside traditional CPU and GPU elements.

“You would need a big GPU for this,” said Milan Patel from MulticoreWare, speaking at last week’s Embedded World show in Nuremberg. “This is a very low power MPU. We have been able to make it smaller without sacrificing accuracy.”

This technology enhances self-checkout systems for smart retail by using AI and computer vision. It enables accurate product identification through object detection, product verification and barcode scanning. Additionally, the system includes AI-powered shoplifting detection, offering real-time monitoring of suspicious customer behaviour and LLM-driven instant alerts to notify store staff of potential theft.

Through its collaboration with AMD, MulticoreWare harnesses the power of the Ryzen Embedded 8000 processor (Mini PC 4X4) to optimise checkout operations, detect customer behaviour anomalies and create a secure, AI-driven shopping ecosystem.

“We are excited to be working closely with AMD in driving the future of AI-powered smart retail,” said Raj Gupta, vice president at MulticoreWare. “Together, we are demonstrating how edge processing enables real-time AI-inferencing for smarter, faster and low-latency retail automation. This relationship paves the way for cutting-edge innovations that will transform smart retail experiences globally.”

Yingyu Xia, vision lead at AMD (www.amd.com), added:”The retail industry is evolving with AI to solve unmet challenges such as increasing store revenue, reducing worker fatigue and enhancing customer experience. MulitcoreWare was able to leverage the multiple compute engines from AMD Ryzen Embedded 8000 processors to assign multi-AI models efficiently, processing multi-sensor data in real time, all in one chip, to deliver a streamlined and cost-effective option for the retail market.”

MulticoreWare leverages AR, data analytics and real-time tracking to create personalised shopping experiences, reduce theft and optimise inventory management. Its products enable dynamic pricing, automated checkouts to reduce wait times, and enhanced in-store security through RFID and camera systems.

MulticoreWare (www.multicorewareinc.com) delivers software IP and engineering services with hardware platform compilers and toolchains, SDK libraries, video codecs, and algorithm and data engineering using vision and non-vision radar, lidar, IMU and GPS sensors. Its products are used in automotive, surveillance, military, smart health, IoT, industrial, robotics and smart cities.