∇·F π(u|y) Σ
Senior DevTech Engineer · NVIDIA

Clement
Etienam
Ph.D.

GPU-accelerated reservoir simulation, physics-informed neural operators, and Bayesian inverse problems. Building the next generation of scientific machine learning for subsurface flow and carbon capture.

Physics Informed neural Operator HPC Reservoir Simulation CCUS Inverse Problems CUDA Machine Learning Gaussian process Sparse Linear Algebra
Clement Etienam
159
Citations
01 —

About


I am a Senior DevTech Engineer in the Energy team at NVIDIA, working at the intersection of GPU computing, scientific machine learning, and subsurface physics. My research focuses on making physics-informed neural operators fast and accurate enough to replace traditional numerical simulators in real-world energy workflows — from reservoir management, reservoir simulation, reservoir history matching to carbon capture and storage at scale.

My PhD at the University of Manchester (main supervisor Rossmary Villegas, with co-supervisors Oliver Dorn, Masoud Babaei, and Constantinos Theodoropoulos, and a postdoc under K.J.H. Law, joint between the University of Manchester and Oak Ridge National Laboratory) focused on reservoir history matching, inverse problems, and parametrising unknown geological fields with exotic generative priors — variational autoencoders (VAE), the discrete cosine transform (DCT), and K-SVD dictionary learning — to regularise otherwise severely ill-posed inversions. The central challenge: given noisy observations y, recover the unknown parameter field u by characterising the posterior

with the field itself reparametrized through a low-dimensional generative code,

where D is a decoder learned via VAE, DCT, or K-SVD, and G : uy is the forward operator — classically a full black-oil PDE solve, now replaced by a learned neural surrogate. This work is applied directly to reservoir history matching, CO₂ plume migration, and ensemble-based inversion (ES-MDA, aREKI) with generative priors.

During my postdoc, I developed the CCR (Cluster Classify Regress) framework for learning highly nonlinear discontinuous functions across sharp phase boundaries — a persistent failure mode for standard regression. Building on this, I have applied PINO and FNO-based surrogates to the Norne field black-oil benchmark (46×112×22 grid, multi-phase, multi-well), solving the parametric PDE family

across thousands of permeability realisations in a single forward pass, achieving up to 6000× speedup over conventional simulators. I contribute to the NVIDIA PhysicsNeMo open-source framework.

6000×
Simulation speedup
159
Total citations
10+
Publications
10yr
Research career
02 —

Education


Postdoctoral Research
Applied Mathematics
2018 — 2020
Developed the CCR (Cluster Classify Regress) framework for learning highly nonlinear, discontinuous functions across sharp phase boundaries — a persistent failure mode for standard regression methods. Supervised by K.J.H. Law.
Ph.D. Engineering Mathematics
2015 — 2018
Reservoir history matching and inverse problems, parametrising unknown geological fields with exotic generative priors — VAE, DCT, and K-SVD dictionary learning — main supervisor Rossmary Villegas, with co-supervisors Oliver Dorn, Masoud Babaei, and Constantinos Theodoropoulos.
Read the full thesis (PDF) →
M.Sc. Petroleum Engineering
2013 — 2014
M.Sc. Thesis: dynamic modelling for haptic feedback in well drilling operations — supervised by Wei Kung-Fung. First formal exposure to dynamical systems and feedback control, ahead of the applied mathematics that followed.
B.Sc. Chemical Engineering
2005 — 2010
Process systems, thermodynamics, and transport phenomena — first exposure to the conservation laws that underpin every reservoir and flow model since.
03 —

Research


2026
Sequential Physics-Constrained Neural Operator Forward Modeling for the Norne Reservoir System
C. Etienam, Y. Juntao, O. Ovcharenko, N. Luiken, T. Onishi, N. Moridis, I. Said
arXiv:2605.28909
2024
Reservoir History Matching of the Norne Field with Generative Exotic Priors and a Coupled Mixture of Experts — PINO Forward Model
C. Etienam, Y. Juntao, O. Ovcharenko, I. Said
arXiv:2406.00889
2025
Accelerating Porous Media Flow Simulations with Fourier Neural Operators: An Application to Geologic Storage of CO₂
A. Chandra, M. Koch, S. Pawar, A. Panda, K. Azizzadenesheli, J. Snippe, F. O. Alpak, F. Hariri, C. Etienam, P. Devarakota, A. Anandkumar, D. Hohl
★ Advanced Theory and Simulations, Wiley · 2025
2024
A Novel AI-Enhanced Reservoir Characterization with a Combined Mixture of Experts — NVIDIA Modulus-based PINO Forward Model
C. Etienam, Y. Juntao, I. Said, O. Ovcharenko, K. Tangsali, P. Dimitrov, K. Hester
arXiv:2404.14447
2022
LIPS — Learning Industrial Physical Simulation Benchmark Suite
M. Leyli-Abadi, A. Marot, J. Picault, D. Danan, M. Yagoubi, B. Donnot, S. Attoui, P. Dimitrov, A. Farjallah, C. Etienam
★ NeurIPS 2022
2020
Ultra-fast Deep Mixtures of Gaussian Process Experts
C. Etienam, K. J. H. Law, S. Wade
arXiv:2006.13309
2019
CCR: Cluster Classify Regress — A General Method for Learning Discontinuous Functions
D. E. Bernholdt, M. R. Cianciosa, C. Etienam, D. L. Green, K. J. H. Law, J. M. Park
★ AIMS Foundations of Data Science · 2019
2019
4D Seismic History Matching Incorporating Unsupervised Learning
C. Etienam
arXiv:1905.07469
04 —

Code


05 —

Connect