Embedded Computing Design

Embedded Computing Design Staff

January 5, 2026

Equipment downtime in a manufacturing facility, or in almost any industrial environment, represents more than just an inconvenience. It is a direct hit to a vendor’s productivity, profitability, and brand. As production lines become more automated and capital-intensive, the tolerance for unexpected failures continues to shrink. That reality is driving manufacturers toward predictive maintenance, where AI and machine learning are used to detect failures long before they result in unplanned shutdowns.

Gary Lin

December 29, 2025

The Secret AI Trio to Make Devices Smarter

System designers increasingly look for platforms capable of handling multiple AI workloads with efficient power consumption and low latency. Recent developments in embedded processors now combine general-purpose compute, graphics acceleration, and dedicated neural processing capabilities into a single System-on-Chip.

This article provides a technical walk-through demonstrating multi-model AI inference performance on the AIMB-2210 Mini-ITX platform featuring the AMD Ryzen Embedded 8000 processor.

Ken Briodagh

December 22, 2025

Video and computer vision are the most impactful applications for AI currently in the embedded industry. Manufacturing, Warehousing and Robotics are using computer vision and video in almost every modern use case, while Smart City Safety and Security is almost entirely focused on video data analysis. And these vertical markets need the AI in the embedded computing at the edge to power their applications and use cases.

Tiera Oliver

December 18, 2025

Crowd behavior and movements can be unpredictable. It’s important that areas that host large quantities of people, such as stadiums, concert venues, and convention centers, are capable of keeping individuals safe in real-time with high-performance analytics.

Ken Briodagh

December 11, 2025

One of the toughest challenges with solving predictive maintenance, enabled by AI, is that there is usually no data available for training the system. Modern equipment may not break down for 5 years or more so there is insufficient data when the product is first launched on when and how it will break down. But AI systems using traditional methods such as Deep Learning can only operate when sufficiently large volumes of data are available for training the system and hence, they are incapable of addressing this problem.