My name is Grigori Fursin. I am the CTO and co-founder of dividiti,
Chief Scientist of the cTuning Foundation (non-profit research organization),
a founding member of the
ACM Task Force on Data, Software, and Reproducibility in Publication,
and the reproducible research evangelist.
I have an interdisciplinary background in computer engineering, physics, neural networks,
quantum electronics and machine learning with a PhD in computer science from the
University of Edinburgh.
I lead research and development for an open Collective Knowledge platform (CK)
to co-design efficient software, hardware and model stack for emerging workloads including deep learning and quantum computing
in terms of speed, accuracy, energy and various costs across diverse data sets and platforms from IoT to datac centers and exacsale supercomputers.
CK platform helps our partners (leading companies and universities) survive
in a Cambrian AI/SW/HW explosion,
quickly prototype and automate AI experiments from shared and optimized components available in the CK repository similar to LEGO(tm),
perform competitive analysis of new AI solutions, reduce all R&D costs, accelerate AI rsearch, and enable efficient brain-inspired computing.
[ My R&D background and short biography ]
||English - fluent (British citizen); Russian - native; French (spoken) - intermediate
||I currently live in Paris suburbs and travel regularly to the UK and USA where I have main projects. However, I am fine to relocate for new projects if it also suits my family.
- 2015-cur.: Co-founder and CTO, dividiti, UK.
- 2014-cur.: Chief Scientist and Technologist, cTuning foundation, France, France.
- 2012-2014: Tenured Research Scientist (associate professor) at INRIA, France.
- 2010-2011: Head of application characterization and optimization group at Intel Exascale Lab, France.
- 2010-cur.: Consultant and scientific advisor (knowledge management, big data predictive analytics, brain-inspired computing,
machine learning based performance/energy/size/cost autotuning, run-time adaptation, SW/HW co-design).
- 2008-cur.: Founder of cTuning.org to crowdsource machine-learning based autotuning as an open research.
- 2007-2010: Guest lecturer at the University of Paris-Sud, France.
- 2007-2010: Tenured Research Scientist (assistant professor) at INRIA, France.
- 2000-2006: Research Associate at the University of Edinburgh, UK.
- 1999-cur.: Evangelist of a collaborative and reproducible research and experimentation in computer engineering.
- 1993-1999: Research Assistant at MIPT, Russia.
- 1992-1993: Founder and CTO of a software startup in Moscow.
- 2004: PhD in computer science with ORS award from the University of Edinburgh, UK.
- 1999: MS in computer engineering with golden medal (summa cum laude) from Moscow Insitute of Physics and Technology, Russia.
- 1997: BS in electronics, mathematics and machine learning (summa cum laude) from Moscow Institute of Physics and Technology, Russia.
||See academic and industrial partners of my cKnowledge.org project
- 2018: Helped initiate 1st open ML-systems tournament on reproducible and Pareto-efficient SW/HW co-design for deep learning (speed, accuracy, costs) at ACM ASPLOS
- 2017: Received ACM CGO test of time award for my research on machine-learning based optimization (CGO'07 publication).
- 2012-2016: Received INRIA award and fellowship for "making an outstanding contribution to research".
- 2014-2015: Received EU TETRACOM grant to develop 4th version of a univeral machine-learning based autotuning framework and public repository for artifact sharing (Collective Knowledge).
- 2012-2014: Developed 3rd version of a universal machine-learning based pluginized autotuning framework supporting multiple objectives including performance,energy,size and cost for a variety of kernels, codelets and large applications with OpenCL, CUDA, OpenMP, and MPI.
- 2014-cur.: Initiated Artifact Evaluation for CGO, PPoPP, ADAPT and PACT (follow-up to my initiative on collaborative and reproducible research).
- 2008-cur.: Established cTuning.org community-driven portal and non-profit foundation to start sharing artifacts along with publications while reusing them to crowdsource software/hardware optimization and combine it with machine learning.
- 2007-cur.: Transferred developed technology to industry and production tools such as mainline GCC; consulted major companies on systematic and reproducible program and architecture performance tuning, run-time adaptation and co-design.
- 2007-2010: Prepared and tought guest MS course on machine learning based optimization and run-time adaptation at the University of Paris-Sud, France.
- 2006-2009: Led research and development of the machine-learning based self-tuning compiler (proposed to crowdsource plugin-based autotuning and combine it with predictive analytics and collective intelligence) in EU FP6 MILEPOST project considered by IBM to be the first in the world.
- 1999-2000: Led research and development of a polyhedral source-to-source compiler together with collaborative plugin-based autotuning infrastructure and repository for memory hierarchy optimization in supercomputers within the EU MHAOTEU project.
- 1999-2006: Prepared foundations for big data driven and machine learning based optimization, run-time adaptation and co-design of computer systems.
- 1998-cur.: Started designing infrastructure and repository for crowdsourcing experiments and sharing results (code, data, models, interactive graphs) in a reproducible way among colleagues and workgroups.
- 1993-1998: Designed novel semiconductor neural network accelerators for a possible brain-inspired computer (served as a motivator for machine-learning based autotuning and collaborative R&D).
|Main techical knowledge (continuously acquire new ones):
Linux, Windows, Android, Python, scikit, neural networks, decision trees, SVM, agile development, large-scale project management, APIs, GCC, LLVM, polyhedral optimizations, ARM compilers, Intel Compilers, Intel VTUNE, C, C++, Java, Fortran, Basic, GPU, OpenCL, CUDA, MPI, OpenMP, PHP, R, MySQL, FPGAs, ElasticSearch, Hadoop, Jenkins, html, apache2, mediawiki, drupal, OpenOffice, Eclipse, SVN, GIT, GIMP2, Adobe Photoshop, Visual Studio, Microsoft Office, Android Studio
|Full academic CV:
||ACM, HiPEAC, IEEE
|My favourite story about Rutherford and a student
||traveling, discovering new cultures, gardening, active sport (football, skiing, swimming, snorkeling, climbing, jogging, ...), photography, reading
||Grigori.Fursin@cTuning.org or email@example.com
Quick access to full academic CV:
"It is not the strongest of the species that survives, or the most intelligent;
it is the one most capable of change" - attributed to Charles Darwin.
(see our ADAPT workshop)
[ All success stories ]
Our CGO'07 paper received "test of time award" at CGO'17!
ARM and dividiti
made a press-release about my Collective Knowledge Technology
[ PDF (page 17) ]
I started crowd-tuning campaign to crowd-benchmark deep learning algorithms
and crowdsource GCC/LLVM tuning while combining it with active learning across diverse data sets
and platforms including mobile devices and cloud services provided
by volunteers using CK framework. You can see latest crowd-results
in our live repository!
- We successfully initiated community-driven pre-reviewing and validation of publications and artifacts
for workshops and conferences - see ADAPT 2016 workshop.
- After sharing all my artifacts and promoting collaborative research since 2007,
I helped initiate Artifact Evaluation for PPoPP, CGO, PACT, RTSS and SC conferences.
Here is our motivation:
- Ed Plowman (director of performance analysis strategy in ARM) suggests to contribute
to Collective Knowledge
and workload automation
[ Slides ]!
- I have received HiPEAC technology transfer award
for my novel Collective Knowledge framework..
- My cTuning technology was referenced by Fujitsu
as closely related to their long-term initiative on "big
data" driven optimization of Exascale computer systems
- My cTuning technology demonstrated the possibility
to fully automate construction of compiler optimization
heuristics for multi-core reconfigurable systems using
machine learning and crowdsourcing - it is considered
by IBM to be the first in the world (2009)
- I extended cTuning-based technology
to develop customized "in house" repository of knowledge
when helping to establish Intel Exascale Lab in France
- My plugin-based compiler interface and technique
to enable autotuning and run-time adaptation for statically compiled
programs was added to mainline GCC (4.6+ and 4.8+
respectively) sponsored by Google [ Details ] (2008-2010)