SWIM AI emerged from stealth mode on April 4, bringing with it a new approach to enable machine learning and data insights without the need for big data processing in the cloud.
SWIM, which stands for Software In Motion, also announced that Simon Crosby has joined the company as CTO. Crosby was a co-founder and CTO at virtualization vendor XenSource, which was acquired by Citrix in 2017 and was also a co-founder and CTO at security vendor Bromium from 2011 until joining SWIM. As a company, SWIM.AI has raised a total of $7 million in funding to date.
"SWIM has developed technologies that allows organizations to analyze and learn from streaming data at the edge of the network in real-time." Crosby told eWEEK.
Crosby noted that the common process for getting streaming data insights today is with edge devices streaming data to some form of cloud big data service, which then parses the data for analysis. In contrast, Crosby said that the SWIM EDX platform consumes data while discovering entities in the data that can be analyzed. The model is continuously self-learning from ingested data to further refine and improve data prediction modelling on edge devices.
"So we're bypassing the whole process of Big Data modelling and learning with SWIM," Crosby said. "We do all the analysis, on the fly, from the data directly."
There are multiple use-cases for edge of network deep learning, including enabling improved networking traffic as well even helping with automobile traffic in the physical world. SWIM.AI along with partner Trafficware announced on April 4 the TidalWave live streaming traffic to help provide insight on motor vehicle traffic.
"TidalWave predicts the state of intersections in major U.S. cities," Crosby said.
The TidalWave service gets its data from sensors deployed at intersections and streets, and is continuously analyzing data to make predictions. Crosby said that the analysis of the data is made available as an API for endpoints, such as autonomous vehicles.
"So you could route a self-driving car through a city and never see a red light," Crosby said.
With many data analysis platforms today, some form of streaming technology such as Apache Kafka or Amazon Kinesis is used to collect data. Crosby explained that the SWIM model is very different consuming data and deriving insights from the data as close as possible to the edge.
"The whole point is that you can bypass the need to send vast amounts of data around by simply learning at the edge where the data is produced," Crosby said.
Crosby explained that the SWIM model is stateful and resilient and doesn't require large data volumes of storage either. He added that the SWIM technology can run on low-power, low-cost ARM devices.
"SWIM builds a distributed fabric across all the compute entities at the edge and then the fabric is effectively an edge data cloud," Crosby said.
With the edge data cloud, SWIM uses an adaptation of the machine learning concept known as convolutional deep neural network to understand data. Crosby explained that the approach enables the model to learn from every data point and is continuously refined.
"We've all been suckered into the idea that cloud is where all the learning happens, but that's not going to be true," Crosby said. "Effectively getting data from the edge to the cloud, cleaning the data and then doing analysis is just too expensive."
Sean Michael Kerner is a senior editor at eWEEK and InternetNews.com. Follow him on Twitter @TechJournalist.