DeepCube

DeepCube developed software solutions to accelerate deep learning on edge devices and data centers. The company's proprietary framework could be deployed on existing hardware, promising significant improvements in speed and memory reduction for deep learning applications. DeepCube's technology was designed to enable advanced AI systems to operate without requiring a connection to the cloud, allowing for efficient and cost-effective implementation of neural networks on edge devices.

The company's inference accelerator applied neural network training techniques, similar to those used in photo and speech recognition advancements, to various applications including manufacturing. DeepCube's system utilized multiple sensors capable of detecting defects too small for the human eye and employed AI-driven decision-making to correct printing errors in real-time. The company claimed its proprietary algorithms could increase data analysis speeds tenfold, positioning it as a unique hardware performance accelerator.

DeepCube's approach aimed to overcome infrastructure, energy, and memory limitations of previous AI models, enabling low-cost deployment. The technology was applicable to various scenarios, such as enabling surveillance cameras to conduct real-time image analysis without an internet connection, allowing autonomous drones to make decisions without connectivity, and facilitating autonomous car operations without cloud reliance.

In October 2021, DeepCube was acquired by Nano Dimension, integrating its deep learning capabilities into Nano Dimension's operations.