diff options
Diffstat (limited to 'bio.tex')
| -rw-r--r-- | bio.tex | 34 |
1 files changed, 34 insertions, 0 deletions
@@ -0,0 +1,34 @@ +\documentclass[10pt]{article} +\usepackage{fontspec} +\usepackage{csquotes} +\usepackage{polyglossia} +\setmainlanguage{english} +\usepackage[backend=biber,style=ieee,maxnames=8]{biblatex} +\renewcommand*{\bibfont}{\small} +\usepackage{titling} +%\renewcommand*{\bibfont}{\footnotesize} + +\addbibresource{bib/csa.bib} + +\title{Suren A. Chilingaryan} +\preauthor{} +\postauthor{} +\author{} +\predate{} +\postdate{} +\date{} +%\posttitle{\par\end{center}} +\setlength{\droptitle}{-4em} + +\begin{document} +\maketitle + + I am specializing in the domain of high-performance and heterogeneous computing, computer architectures, and parallel algorithms. On a technical side, I have experience in performance analysis and optimization, parallel programming, low-latency communication, and cloud platforms. Working at Institute of Data Processing and Electronics (IPE) at KIT, I apply these technologies to build software instrumentation for distributed data acquisition and control systems. + + I built and are currently maintaining a highly-available cloud platform for KATRIN (KArlsruhe TRItium Neutrino) data acquisition and slow-control systems~\cite{katrin2015detector,katrin2018first}. Parts of the system are adapted to support ASEC (Aragats Space Environmental Center) and SEVAN (Space Environmental Viewing and Analysis Network) particle detector networks in Armenia to study thunderstorm phenomena~\cite{csa2009sevan, chili2010thunderstorm}. I led a software work package of the UFO (Ultra Fast tOmography) project aimed to build a novel instrumentation for high-speed synchrotron imaging with online reconstruction and image-based feedback loop~\cite{kopmann2017ufo}. We developed a control system integrating the beam line devices with a GPU-based image-processing cluster and steering the data from the cameras until the storage~\cite{stevanovic2015concert}. IPE is actively designing novel electronics for multiple collaborations~\cite{caselle2013camera,caselle2014kapture}. Our group is looking for a hybrid solutions coupling the high-speed electronics with fast, but flexible software running on GPUs and other parallel accelerators~\cite{vogelgesang2016dgma}. For instance, recently we have performed a case-study aimed to evaluate the possibility of building the next generation of CMS track trigger using GPUs with round-trip latency below 6 us~\cite{mohr2017cms}. + + It is a challenging task to design an efficient computing system. Well designed data flow, a hierarchy of intelligent caches, and efficient parallel algorithms can drastically reduce required investments. Throughout all projects, we take a holistic approach to understand project requirements, identify bottlenecks, and optimize performance-critical components. To get an in-depth understanding of available parallel architectures, I have systematically applied micro-benchmarking techniques. It allowed to find multiple undocumented properties of the available hardware and to develop a range of techniques to balance the load between different computational and memory units achieving higher hardware utilization~\cite{csa2018sbac}. We have developed a pipelined image processing framework and contributed parallel algorithms addressing various hardware platforms including IBM Power, Intel Xeon Phi, and multiple GPU architectures~\cite{csa2011pyhst,vogelgesang2012ufo,ashkarin2015,rshkarin2015,cavadini2018upiv}. To enable interactive remote visualization of large tomographic volumes, we develop a web-based visualization framework combining client- and server-side rendering techniques~\cite{ntj2017wave}. Because of the client-side component, high interactivity is achieved with only small investments in the data center hardware. On the other hand, the server-side component allows to improve quality on demand and makes visualization possible also for slow hand-held devices. + + +\printbibliography +\end{document} |