DataCenterNews Asia Pacific - Specialist news for cloud & data center decision-makers
Story image
Mellanox teams with machine learning firm to advance speech technology
Fri, 16th Dec 2016
FYI, this story is more than a year old

Mellanox is now the chosen provider of Ethernet solutions for a next-generation machine learning center that aims to achieve high speech recognition rates of 97%.

The company has partnered with iFLYTEK, a Chinese intelligent speech and language technology firm to deploy end-to-end 25G and 100G Ethernet solutions based on ConnectX adapters and Spectrum switches for the machine learning center.

iFLYTEK executive vice president Zhiguo Wang says that the Mellanox solutions enabled the next-gen machine learning center to boost performance and provide for the future.

“Moreover, we leverage the scalability of Mellanox Ethernet solutions to grow our compute and storage needs in the most efficient manner," Wang says.

iFLYTEK has been dedicated to the research of speech and language technologies, provision of speech information services, and integration of E-government systems. The company says its data center needs demand high-performance, efficient, scalable and compatible networking solutions.

“Speech recognition has made tremendous advancements in the past couple of years and we are excited to partner with iFLYTEK to build the next generation of machine learning systems for this technology development,” says Gilad Shainer, vice president of marketing at Mellanox Technologies.

Mellanox states that machine learnings can handle massive concurrent traffic, and is able to support unpredictable future business growth. In order to do that, Mellanox will provide automatic network provision and management, convergence network infrastructure with Quality of Service and RDMA over Ethernet.

“Mellanox's 25G/100G Ethernet solutions provide the most cost effective, high performance Ethernet connectivity allowing iFLYTEK to provide the needed compute and storage resources to their developers," Shainer concludes.