Qualcomm Technologies, Inc. is acquiring Edge Impulse, a startup that assists developers in running artificial intelligence models on connected devices.
The two companies announced today that they have reached an agreement, though financial terms were not disclosed.
Edge Impulse, based in San Jose, California, raised $54 million from investors prior to the acquisition. It offers a cloud platform that businesses can use to train AI models optimized for connected devices. According to Edge Impulse, its software is used by over 170,000 developers worldwide.
AI models optimized to run at the edge are often trained on sensor data. For example, a neural network designed to detect overheating in industrial equipment might be trained on temperature measurement data from production lines. Edge Impulse provides capabilities to transform these measurements into AI training datasets.
The company's platform includes tools that AI teams can use to create data pipelines. These tools are automated software workflows for assembling training datasets from sensor readings. For instance, these workflows can filter out duplicate and erroneous records.
Edge Impulse’s platform also offers tools for performing feature engineering. This is the task of condensing raw sensor data into a form more manageable for AI models. For example, developers might convert a set of temperature readings into a single average value that is easier to analyze.
After software teams create training datasets, Edge Impulse helps identify suitable AI model architectures for projects. Connected devices often have limited processing power. Edge Impulse displays hardware requirements for different AI architectures, making it easier to find ones efficient enough to run on a company's connected devices.
After developers use Edge Impulse to train custom AI models, the platform packages the algorithms into C++ libraries. C++ is a more hardware-efficient programming language compared to Python, which is the preferred language for AI development. The Edge platform claims it can reduce memory usage of neural networks by over 60%.
The company offers its development platform along with a device called BrickML. This compact rectangular computing module is designed to run AI models. Companies can attach it to industrial equipment, collect sensor readings, and use the onboard neural network to analyze data locally.
"Edge Impulse provides developers with a tool that automates data collection, simplifies model training, provides advanced optimization tools, and offers one-click deployment to multiple hardware platforms," wrote co-founder and CEO Zach Shelby in a blog post.
Qualcomm's acquisition of Edge Impulse comes just weeks after the chipmaker launched Dragonwing, a processor portfolio optimized for connected devices. Some chips in the lineup include integrated graphics processing units for running AI models. Others feature reliability optimizations allowing them to operate in sub-zero temperatures.
Currently, Edge Impulse supports two Dragonwing processors: QCS6490 and QCS5430. Both chips can be used to power rugged mobile devices, while the latter module is also suitable for robotics projects. After being acquired by Qualcomm, Edge Impulse plans to expand support for more processors in the Dragonwing portfolio.
"In addition to supporting Qualcomm Technologies' hardware, we will continue to support edge hardware from our broad partner base, including MCUs, CPUs, GPUs, and NPUs," Shelby elaborated.
