Guillaume Barnier's page

GUILLAUME BARNIER

Phone: +1-720-838-1835 • Email: barnier@gmail.com • Website: gbarnier.github.io • GitHub: github.com/gbarnier

SUMMARY

> Expert in machine learning, scientific computing, and optimization of large-scale nonlinear inverse problems
> Strong interest and knowledge in reinforcement learning algorithms
> Excellent high-performance computing (HPC) skills with deep programming knowledge of graphics processing units (GPU) in C++/CUDA
> Won the 2019 “Best Paper Presented by a Student at the Annual Meeting” award from the Society of Exploration Geophysicists (SEG)

EDUCATION

Stanford University, Ph.D. in Geophysics (GPA: 3.960) - Exp. 2021
Colorado School of Mines, MSc in Geophysics (GPA: 4.000) - 2013
Télécom Paris University, MSc in Electrical Engineering (GPA: 3.600) - 2007

MACHINE LEARNING SKILLS

ML coursework: Reinforcement Learning (CS 234), Deep Learning (CS 230), and Machine Learning (CS 229)
Mathematics coursework: Stochastic Processes (MATH 136) and Stochastic Differential Equations (MATH 236)
Programming: C++, CUDA, Fortran, pybind11, Unix, MPI, Dask, HTML, CSS
Deep Learning Frameworks: Python, PyTorch, NumPy, some use of Keras and TensorFlow

RESEARCH EXPERIENCE

Deep Learning, Computer Vision, and Healthcare

Radiology and Neuroimaging Department, Stanford School of Medicine (Jan. 2021 - Present)
Research assistant (Prof. E. Tong)

> Initiated and led a project to predict brain tissue damage for ischemic stroke patients using deep supervised learning based on convolutional neural nets (CNN) and 4D computed tomography perfusion (CTP) data
> Developed an approach to bypass the use of costly and potentially unreliable commercial software packages
> Method may have profound impact on stroke imaging by improving diagnostic reliability and making CTP technique more widely available and equitable

Geophysical Seismic Imaging

Stanford Exploration Project, Stanford Department of Geophysics (Sep. 2013 - Present)
Doctoral Researcher (Prof. B. Biondi)

> Conceived innovative industry-impactful algorithm to produce accurate images of the Earth’s subsurface using seismic waves
> Created a new robust gradient-descent optimization scheme combined with the variable projection method to mitigate the presence of spurious local minima inherent to full waveform inversion (FWI), a challenging high-dimension prominent seismology problem
> Developed an industry-quality finite-difference code implemented in C++/CUDA using GPUs to enhance computational efficiency for large-scale 3D field applications (tens of terabytes of data)
> Successfully applied Ph.D. algorithm on a deep-water node-acquisition 3D field dataset from the Gulf of Mexico
> Co-led a team of students for the development of an advanced Python optimization library for solving large-scale inverse problems

Numerical Methods for Wave Propagation

Center for Induced and Triggered Seismicity, Stanford Department of Geophysics (Jan. 2015 - Jan. 2017)
Research assistant (Prof. E. Dunham)

> Developed an early-warning tsunami modeling and prediction algorithm using a data assimilation technique with Kalman filters
> Applied method to a 2D tsunami model based on dynamic rupture simulations of the 2011 Tohoku earthquake
> Accurately reconstructed tsunami wavefield prior to wave arrival at the coast

INDUSTRY EXPERIENCE

Self-Initiated Industry Partnerships

Nov. 2018 - Jan. 2021
Seismic Imaging Research Consultant

> Successfully approached and convinced management of three major oil companies to invest in personal imaging algorithm and deployed solution on companies’ computing infrastructures

BP America - Houston, TX, USA

Summers 2016 and 2017
Seismic Imaging Research Scientist

> Provided research team with expertise on personal Ph.D. project
> Deployed algorithm on BP’s HPC system
> Showed value and potential business impact for improving images in complex geological settings

Chevron - San Ramon, CA, USA

Jun. 2015 - Sep. 2015
Seismic Imaging Research Scientist

> Improved and applied a Bayesian geophysical uncertainty estimation technique based on the Metropolis-Hastings for oil reserve quantification

Total - Pau, France

May 2012 - Sep. 2012
Seismic Imaging Research Scientist

> Analyzed and identified an optimal reservoir-characterization solution (among three software packages) adequate to the team’s needs and budget, and convinced management committee to follow selection recommendation

J.P. Morgan - London, UK

Jun 2007 - May 2010
Fixed Income Hedge Fund Sales and Structurer

> Provided pricing, trade executions, and financial advice on interest rates derivative products to major investment firms including Pacific Investment Management Company (PIMCO)
> Took the initiative to develop a novel interest-rate swaps modeling code with improved flexibility and accuracy (compared to available commercial package), which became the team’s primary tool for asset pricing

AWARDS AND ACHIEVEMENTS

Academics

2019 Award for Best Paper Presented by a Student at the Society of Exploration Geophysicists (SEG)
2013 Colorado School of Mines Mendenhall award for outstanding academic achievement
2012 Colorado School of Mines Hess Corp Fellowship

Trail-running

2018 First place overall, Spartan Race Diablo Grande Super Open
2018 First place overall, Inside Trail Pacifica Foothills Trail Half Marathon
2015 First place overall, Mount Diablo Coastal Trail Runs Half Marathon

LANGUAGES

French, English, Spanish (fluent), Hebrew (beginner), Bahasa Indonesia (survival level)

PUBLICATIONS

  1. E. Biondi, G. Barnier, R. G. Clapp, F. Picetti, S. Farris, An object-oriented optimization framework for large-scale inverse problems: Computers & Geosciences (submitted).
  2. Barnier, G. and Biondi, E., 2020. Full waveform inversion by model extension using a model-space multi-scale approach. In SEG Technical Program Expanded Abstracts 2020 (pp. 646-650). Society of Exploration Geophysicists.
  3. Biondi, E. and Barnier, G., 2020. Elastic-parameter estimation by combining full-waveform inversion by model extension and target-oriented elastic inversion. In SEG Technical Program Expanded Abstracts 2020 (pp. 735-739). Society of Exploration Geophysicists.
  4. Barnier, G., Biondi, E. and Clapp, R., 2019. Waveform inversion by model reduction using spline interpolation. In SEG Technical Program Expanded Abstracts 2019 (pp. 1400-1404). Society of Exploration Geophysicists. Won award for Best Paper Presented by a Student.
  5. Yang, Y., Dunham, E.M., Barnier, G. and Almquist, M., 2019. Tsunami wavefield reconstruction and forecasting using the ensemble Kalman filter. Geophysical Research Letters, 46(2), pp.853-860.
  6. Barnier, G., Biondi, E. and Biondi, B., 2018. Full waveform inversion by model extension. In SEG Technical Program Expanded Abstracts 2018 (pp. 1183-1187). Society of Exploration Geophysicists.
  7. Biondi, E., Biondi, B. and Barnier, G., 2018. Target-oriented elastic full-waveform inversion through extended-migration redatuming. In SEG Technical Program Expanded Abstracts 2018 (pp. 1228-1232). Society of Exploration Geophysicists.
  8. Barnier, G., Biondi, E. and Biondi, B., 2018, June. A Modified Approach for Tomographic Full Waveform Inversion Using Variable Projection. In 80th EAGE Conference and Exhibition 2018 (Vol. 2018, No. 1, pp. 1-5). European Association of Geoscientists & Engineers.
  9. G. Barnier, E. Biondi, and B. Biondi, 2017, A modified approach to tomographic full-waveform inversion: SIAM Imaging Symposium, Rice University, Houston
  10. Biondi, E., Barnier, G. and Biondi, B., 2017. Preconditioned elastic full-waveform inversion with approximated Hessian. In SEG Technical Program Expanded Abstracts 2017 (pp. 1654-1658). Society of Exploration Geophysicists.
  11. Barnier, G. and Dunham, E.M., 2016, December. Tsunami Modeling and Prediction Using a Data Assimilation Technique with Kalman Filters. In AGU Fall Meeting Abstracts (Vol. 2016, pp. NH41A-1754).
  12. Revil, A., Barnier, G., Sava, P.C., Jardani, A. and Kulessa, B., 2014, December. Seismoelectric coupling in partially water-saturated porous media: From the theory to the detection of saturation fronts. In AGU Fall Meeting Abstracts (Vol. 2014, pp. NS31D-06).
  13. Revil, A., Barnier, G., Karaoulis, M., Sava, P., Jardani, A. and Kulessa, B., 2014. Seismoelectric coupling in unsaturated porous media: theory, petrophysics, and saturation front localization using an electroacoustic approach. Geophysical Journal International, 196(2), pp.867-884.
  14. Barnier, G., 2013. Detecting electrical and hydraulic heterogeneities using seismic focusing and seismoelectric conversions (Doctoral dissertation, Colorado School of Mines).

INVITED CONFERENCE PRESENTATIONS

  1. "Full waveform inversion by model extension using a model-space multi-scale approach" at the 89th SEG Annual meeting 2020, invited speaker at the post-convention workshop "Low-frequency FWI: How low do we need to go?" (Houston, USA)
  2. "Extended imaging for robust velocity-model building and elastic target-oriented full-waveform inversion", invited speaker at the 8th Chevron FWI workshop 2019 (Houston, USA)
  3. "Exploiting the entire data bandwidth by extended imaging for avoiding cycle skipping and elastic target-oriented inversion" at the 88th SEG Annual meeting 2019, invited speaker at the post-convention workshop "Value of High-frequency FWI Models" (San Antonio, USA)
  4. "Full Waveform Inversion by Model Extension", invited speaker at the 2019 SIAM Conference on Mathematical and Computational issue in the Geosciences (Houston, USA)
  5. "Effects of top-salt picking inaccuracies on subsalt image quality" at the 85th SEG Annual meeting 2015, invited speaker at the post-convention workshop "Challenges and Opportunities in Subsalt Imaging" (New Orleans, USA)