Highlights:

  • The platform offers tools for AI teams to build data pipelines, automating dataset creation from sensor readings by filtering duplicates and errors.
  • The company provides its platform with BrickML, a compact module for running AI models on industrial equipment, enabling local sensor data analysis.

Qualcomm Technologies Inc. is acquiring Edge Impulse Inc., a startup specializing in enabling developers to run AI models on connected devices.

The companies recently announced the deal but did not reveal the financial terms.

San Jose, California-based Edge Impulse secured USD 54 million in funding from investors before the acquisition. The company offers a cloud platform that enables businesses to train AI models optimized for connected devices. Edge Impulse reports that its software is utilized by over 170,000 developers globally.

AI models optimized for edge computing are often trained using sensor data. For instance, a neural network designed to detect overheating in industrial equipment might be trained on temperature readings from a production line. Edge Impulse offers tools to convert such data into AI training datasets.

The company’s platform provides tools for AI teams to build data pipelines—automated workflows that streamline the process of creating training datasets from sensor readings. These workflows can, for instance, identify and remove duplicate or erroneous records.

Edge Impulse’s platform also includes tools for feature engineering, which involves transforming raw sensor data into a more AI-friendly format. For example, a developer could condense a set of temperature readings into a single average value, making it easier for the model to process.

Once a software team has built a training dataset, Edge Impulse assists in selecting an AI model architecture that fits the project’s needs. Since connected devices often have limited processing power, the platform showcases the hardware requirements of various AI architectures, helping teams identify models that can run efficiently on their devices.

Once developers train a custom AI model with Edge Impulse, the platform compiles the algorithm into a C library. Compared to Python, the preferred language for AI development, C offers greater hardware efficiency. Edge Impulse claims its platform can reduce neural network memory usage by over 60%.

Alongside its development platform, the company offers BrickML, a compact, rectangular computing module designed for running AI models. Businesses can attach it to industrial equipment to gather sensor data and analyze it locally using built-in neural networks.

“Edge Impulse gives developers a tool that automates data collection, simplifies model training, provides advanced optimization tools, and offers one-click deployment to many types of hardware,” Zach Shelby, Co-founder and Chief Executive Officer wrote in a blog post recently.

Qualcomm’s acquisition of Edge Impulse follows the recent launch of Dragonwing, its processor portfolio designed for connected devices. Some chips in the lineup include an integrated GPU for running AI models, while others feature reliability enhancements that enable operation in sub-freezing temperatures.

Edge Impulse currently supports two Dragonwing processors, the QCS6490 and QCS5430. Both are suitable for ruggedized mobile devices, while the QCS5430 is also well-suited for robotics applications. After joining Qualcomm, Edge Impulse intends to expand support to additional processors within the Dragonwing portfolio.

Shelby detailed, “In addition to support for Qualcomm Technologies’ hardware, we’re continuing support for edge hardware from our wide partner base, including MCUs, CPUs, GPUs and NPUs.”