ETH Zurich Foundation, SnowBell


Mohsen Ewaida

An FPGA Computing Platform for Machine Learning in Data Analytics

Major players in the IT Industry (Intel, Microsoft, AWS, Xilinx) are strongly pushing toward FPGA devices to run machine learning applications. FPGAs have the capability to lower the cost of machine learning IT infrastructure, lower its energy footprint, and meet the stringent service requirements of businesses. However, the complexity of developing applications for FPGAs and the scarce pool of talent represent an obstacle for many businesses that want to use FPGAs.

SnowBell will unlock FPGA platforms for businesses allowing them to run their ML applications on FPGA devices at zero development cost. In SnowBell, they will develop a portable and rich portofolio of ML algorithms for FPGAs. In addition, SnowBell will offer a software environment to seamlessly integrate popular ML frameworks and applications with its portofolio. Customers will have the opportunity to tailor SnowBell ML portofolio to their deployment environment (Cloud, on-premises, embedded) and their problem dimensions (Model size, data structures and layout).