Mark Khait Email & Phone Number

Milan, Lombardy, Italy

Delft Area, Netherlands

2021-2022: Implementation of multiphase compositional thermal flow solver with gravity, capillarity, diffusion, kinetic and equilibrium reactions, precipitation/dissolution (through OBL approach) in GEOSX framework.2019-2021: GPU based compositional simulation using OBL parametrization approach. Study of performance improvement strategies for GPU based general purpose compositional simulation. Demonstration of applicability of GPU based parametrization approach (OBL) for reactive compositional simulations

Delft Area, Netherlands

After spending 11 years in the oil&gas industry, I moved on back to academia to learn more about reservoir modelling on the non-linear level. Inspired by the parametrization approach to aid flash calculations in compositional space (CSP), together with my supervisor Denis Voskov, we developed a novel linearization approach and found out that it improves numerical modelling of complex nonlinear processes in many ways. We demonstrated the applicability of operator-based linearization (OBL) to black-oil, compositional, thermal-compositional and geothermal formulations in a number of peer-reviewed publications (see the links below). OBL changes the perspective and structure of a modelling framework by introducing discretization in physical properties in addition to space and time discretizations. Based on this idea, I have designed and implemented the Delft Advanced Reservoir Terra Simulator (DARTS) platform. The key advantage of its architecture is that the computations of physical properties are detached from the main simulation engines. This provides great flexibility for implementations of various formulations, furthermore enhanced by exposing all necessary interfaces of DARTS in Python. On the other hand, the simulation performance also benefits from the architecture, as the main simulation kernels become simplified. DARTS core functionality is written in C++11\OpenMP\CUDA and provides great performance as the entire simulation can be run on GPU architecture. DARTS platform is now used for general purpose reservoir simulation, inverse modelling, and uncertainty quantification by MSc students, doctoral, and postdoctoral researchers.

Bel Air

Development and implementation of GPU-based Sequential Gaussian Simulation. The results were presented at the Fourth EAGE Workshop on High Performance Computing for Upstream (2019).

Houston, Texas

During this internship, I learned the principles of machine learning and convolutional neural networks (CNN) in particular. Working with Geology Technology Team, I development of a deep learning approach towards carbonate thin section classification according to Dunham textures. Using labelled high-resolution thin section images, CNN with Inception-v3 architecture was fine-tuned for around 140 epochs and accuracy levels of 86% and 83% were achieved on validation and test datasets. In addition, I developed a stand-alone graphical application which allows applying the trained network to an area of interest of any image. We have published the results of our work in Petrophysics Journal (see the link below)

Ufa, Russian Federation

I worked in a small young team of scientific software engineers and mathematicians. We performed a full cycle of development, deployment, and support of a in-house 3-phase black-oil reservoir simulator. I was focused on implementation of parallel versions of the simulator for shared and distributed memory systems, including high-performance implementation of iterative linear solver GMRES, preconditioned by CPR-AMG and ILU(0).

United States

I took the opportunity to participate in a collaboration project between Rosneft Oil Company and Stanford University. During 6 month visit, I learned the structure of General Purpose Reservoir Simulator (GPRS), developed by SUPRI-B group in Earth Sciences, and introduced OpenMP thread parallelism in its linear solver library.

Ufa, Russian Federation

I was introduced to the problems posed by the oil&gas industry in reservoir modelling.In the beginning, I worked on the development of graphical applications for pre- andpost-processing of simulation data (Delphi, Object Pascal). Then, I got an assignment related directly to reservoir simulation - I studied and implemented various algorithmic components of algebraic multigrid solver (AMG). Later, I worked on its parallel version form for shared and distributed memory systems (C++/OpenMP/MPI).

Doctor Of Philosophy - Phd, Petroleum Engineering

M.Sc. In Information Security (Distinction), Computer And Information Systems Security/Information Assurance, 4.98 / 5

Thesis title: "Algorithms and software for a cryptographic system of data protection"