%\subtitle{Research and Technology}{\faSuitcase}
\subtitle{Research and development}{\faSuitcase}

While the focus of my research is computing technolgies, the developed instrumentation enabled major scientific break-troughs achieved by KATRIN~\cite{katrin2019limit} and ASEC~\cite{chili2010thunderstorm} collaborations.
Bellow are referenced selected peer-reviewed publications which are either edited by me and my students or where we made a significant contribution.

\begin{verycomplexevents}
% We don't list experiments which are handled torugh the software (with exceptions dedicated applications?), only direct impact.
% Skipped:
%  buzmakov - just mentions UFO camera. overall not a high ranked paper.
% \subsectiontitle{Technology for the future data acquisition systems}

% ?? make subsections with dates: \eventsection

 \datedeventsection{\since{2011}}{High-bandwidth data acquisition and data-driven control}{}
%  \research	{starting}		{}		{}		{Adoption of}{high-performance and cloud computing for online data processing}
  \research	{\ivl{2018}{2019}}	{PyHST}		{}		{Fine tunning}{tomographic reconstruction to evolving GPU architectures trough performance modelings~\cite{csa2020rtip}}
  \research	{\ivl{2016}{2017}}	{}		{CMS}		{Participated}{in a case study on applications of GPUs in L1 track trigger for the next upgrade of CMS experiment~\cite{mohr2017cms}}
  \technology	{\ivl{2016}{2017}}	{UFO}		{KARA}		{Designed}{a full chain of instrumentation for high-speed synchrotron imaging with online reconstruction and image-based feedback loop~\cite{kopmann2017ufo,stevanovic2015concert}}
  \research	{\ivl{2015}{2016}}	{Alps}		{KARA}		{Researched}{low-latency communication mechanisms for data-driven control applications~\cite{vogelgesang2016dgma}} %dritschler2014using
  \technology	{\ivl{2014}{2015}}	{Alps}		{KARA}		{Implemented}{fast DMA drivers with GPUDirect/DirectGMA support~\cite{rota2015dma}}
  \research	{\ivl{2013}{2014}}	{UFO,PyHST}	{KARA,ESRF}	{Reviewed}{assymptotically fast methods of tomographic reconstruition suitable for GPU architectures~\cite{rshkarin2015}}
  \research	{\ivl{2011}{2013}}	{UFO}		{KARA}		{Researched}{pipelined architectures for online processing of image-streams~\cite{vogelgesang2012ufo}}
  \technology	{\ivl{2011}{2013}}	{Alps}		{KARA}		{Developed}{streaming data acquisition platform for scientific cameras (readout,debugging,storage)~\cite{caselle2013camera}}% readout framework, camera drivers, absraction, streaming storage, scripting & debugging

 \datedeventsection{\since{2007}}{Parallel architectures, performance analysis, and software optimization}{verycomplexevents}
  \technology	{\ivl{2020}{2021}}	{CCPi}		{UoM}		{Applied}{methods of approximate computing to enable reconstruction of large datasets~\cite{ametova2021neutron}}
  \research	{\ivl{2017}{2018}}	{PyHST}		{}		{Researched}{performance inbalances and hidden parallelism in GPU architectures~\cite{csa2018sbac}} % Balancing part is new, everything else was done earlier...
  \technology	{\ivl{2014}{2017}}	{UFO}		{KARA}		{Investigated}{viable compromises between quality and parallelization capabilites of tomographic algorithms~\cite{ashkarin2015}}
  \technology	{\ivl{2013}{2014}}	{UFO}		{}		{Developed}{parallel algorithms for uPIV (micro-particle velocimetry)~\cite{cavadini2018upiv}}
  \technology	{2010}			{MRSES}		{ASEC}		{Leveraged}{PoweXCell architecture for MRSES feature selection algorithm [5000x speed-up]}
  \technology	{\ivl{2009}{2010}}	{PyHST}		{KARA,ESRF}	{Optimized}{PyHST tomographic reconstruction framework~\cite{csa2011pyhst}}
  \technology	{\ivl{2009}{2010}}	{DictHW}	{}		{Developed}{GPU implementation of digital image correlation and tracking algorithm [10x speed-up]}
  \technology	{\ivl{2007}{2008}}	{XMLBench}	{}		{Carried out}{a performance study of open-source XML frameworks~\cite{csa2009xmlbench}}

 \datedeventsection{\since{1999}}{Digitization, data organization, and distributed control systems}{verycomplexevents}
  \research	{\since{2019}}		{KaaS}		{KATRIN}	{Research}{cloud technologies for highly heterogeneous control systems in large-scale scientific experiments [expected]} % KATRIN paper to be added, ADEI-cloudificatin separated as stand-alone point
  \research	{\ivl{2015}{2017}}	{WAVe}		{KARA}		{Researched}{remote visualization techniques for large and time-resolved tomographic volumes~\cite{ntj2017wave,losel2020biomedisa}} % Compression could be continuation of this (19/20)
  \research	{\ivl{2013}{2015}}	{ADEI}		{ASEC}		{Researched}{emerging web technologies for management and visualization of terabyte-scale archives with time-series}	% This is about ADEI2 attempts (but no publication)
  \technology	{\ivl{2011}{2014}}	{ADEI}		{ASEC}		{Converted}{KATRIN data management system into a full flagged platform for time-series exploration and analysis}	% ADEI analysis, etc.
  \technology	{\ivl{2008}{2010}}	{ADEI}		{KATRIN}	{Developed}{data management components of KATRIN slow control system~\cite{csa2010adei}}				% Lets assume it includes ADEI/Control
  \technology	{\ivl{2007}{2008}}	{}		{KATRIN}	{Stabilized}{KATRIN slow control system for production use~\cite{katrin2015detector}}
  \technology	{\ivl{2005}{2006}}	{ADAS}		{ASEC}		{Developed}{a distributed data acquisition system for particle detector networks~\cite{csa2009sevan}}
  \research	{\ivl{2002}{2004}}	{ADAS}		{ASEC,KATRIN}	{Researched}{fast network protocols for heterogeneous slow control systems~\cite{eppler2004opc}}
  \technology	{\ivl{1999}{2001}}	{}		{}		{Evaluated}{hardware-accelerated neural-networks for trigger applications~\cite{vardanyan2001sand}}
\end{verycomplexevents}

%  \research {\since{2020}}		{Group - Application }
%  \research {\ivl{2016}{2017}}	{Group - Supervised a case study on use of GPUs in L1 track trigger for the next upgrade of CMS experiment~\cite{mohr2017cms}.}
%  \technology {\ivl{2011}{2016}}	{Participated in the development of streaming data acquisition platform aimed at control systems with image-based feedback loops and was responsible for the software part~\cite{caselle2013camera,rota2015dma}.}
%  \technology {\ivl{2007}{2009}}	{Developed data management components of KATRIN slow control system~\cite{katrin2015detector}.}
%  \technology {\ivl{2005}{2007}}	{Developed distributed DAQ system for particle detector networks~
%\cite{csa2009sevan}.}
%  \technology {\ivl{2002}{2005}}	{Designed a high-performance extension for OPC XML-DA protocol aimed on slow control systems with high sampling rates.}
%  \technology {\ivl{2020}{2021}}	{Collaborated with scientists from University of Manchester on optimization of their tomography software: suggested methods of approximate computing to enable also reconstruction of large datasets~\cite{ametova2021neutron}.}
%  \research {\ivl{2012}{2019}}	{Analyzed major parallel architectures from AMD, NVIDIA, and Intel using micro-benchmarking techniques. Build a performance model of the back-projection algorithm and fine-tuned the parallel implementation for each architecture accordingly. The measured speed-ups over the state-of-the-art parallel solutions range from 2 to 7 times depending on the considered architecture. Mixed precission.}
%  \technology {\ivl{2011}{2019}}	{A new control-system for high-speed synchrotron imaging with online reconstruction and image-based feedback loop~\cite{kopmann2017ufo}}
%  \research {\ivl{2013}{2017}}	{With a group of 1 PhD and 2 Master students we have investigated iterative reconstruction algorithms to improve reconstruction of noisy or undersampled data. We searched for practical solutions providing viable balance between resulting quality and parallelization capabilities~\cite{ashkarin2015}. The selected algorithms were adapted for the GPU architecture and integrated into the UFO  framework.}
%  \research {\ivl{2013}{2015}}	{Supervised a Master student researching methods of tomographic reconstruction which are both assymtotically faster than traditionally used \emph{Filtered Back Projection} and suitable for execution on GPU architectures~\cite{rshkarin2015}.}
%  \research {\ivl{2011}{2014}}	{With a group of 2 PhD and several Master students we developed a novel architecture for pipelined processing of image-streams which leveraged available parallelism in GPUs (and other massively parallel architectures) and was scaling across multiple GPUs and multi-GPU nodes automatically~\cite{vogelgesang2012ufo}.}
%  \technology {\ivl{2009}{2010}}	{Collaborated with researchers from ESRF facility on high-speed tomographic reconstruction and improved performance of PyHST software by factor of 30~\cite{csa2011pyhst}.}