Sevin Varoglu

Sevin Varoglu is a MLSys Engineer at OctoML. Prior to joining OctoML, Sevin was a Senior Staff Software Engineer at Intel Corporation from October 2008 to May 2021. At Intel, Sevin led a small team of engineers in Application Engineering Group under Habana Labs. Sevin's team was responsible for enabling deep learning models on Habana AI accelerators and driving solutions to customers.

Prior to leading the Application Engineering Group, Sevin worked in the Frameworks and Optimizations team under Artificial Intelligence Products Group (Nervana). In this role, Sevin drove integration of nGraph into popular deep learning frameworks, provided model support and validation, and worked on optimizing inference and training performance bottlenecks.

Sevin's most notable contribution at Intel was delivering a scalable data-parallel distributed training approach for MXNet via nGraph running on CPUs. This approach significantly improved the performance of MXNet-based deep learning models.

In addition to their work in the Artificial Intelligence Products Group, Sevin also worked on application-level binary translation software in Software and Services Group to detect advanced malware attacks with low overhead and high accuracy in Windows Operating Systems.

Before joining Intel, Sevin designed and implemented graphics memory features in Graphics Core Driver Team under Visual Computing Group, focusing on Windows Device Drivers for Intel's integrated graphics products.

Sevin Varoglu has a MS in Electrical and Computer Engineering from UC Irvine, a BS in Computer Engineering from Eastern Mediterranean University, and a PhD in Electrical and Computer Engineering from UC Irvine.