Geomodelling
Current methods for modeling rollfront mineral deposits rely heavily on manual, subjective interpretations that often ignore the complex hydrodynamic and geochemical realities of their formation. To eliminate this human bias and increase accuracy, I developed a rigorous, physics-based numerical model that simulates the genesis of these deposits. By leveraging computational fluid dynamics, Darcy’s Law, and the Law of Mass Action, this reactive transport model accurately reproduces the fluid flows, dissolution, and precipitation of mineral complexes in porous media, providing a deep quantitative understanding of how these epigenetic deposits evolve.
Building on this numerical foundation, I engineered a novel geostatistical method and a dedicated software tool for highly accurate 3D geological modeling and reserve calculations. By integrating Pollock’s algorithm for streamline generation, this approach explicitly honors the hydrodynamic constraints governing fluid flows in ore-bearing layers. The system not only outperforms conventional manual contouring in resource estimation but also serves as a robust engine for generating synthetic deposit data—a critical asset for verifying existing models and training future Physics-Informed Neural Networks.
CFD and Reactive Transport
In-Situ Leaching (ISL) for uranium extraction frequently suffers from operational inefficiencies, such as the formation of stagnant zones and the costly loss of leaching reagents beyond production boundaries due to suboptimal well configurations. To mitigate these physical and economic losses, comprehensive 3D reactive transport models were developed. By coupling mass conservation and Darcy’s Law with the chemical dissolution kinetics of uranium minerals (UO₂ and UO₃), the complex hydrodynamics and acid-rock interactions within heterogeneous strata were accurately simulated. These physics-based numerical models enabled the multi-criteria optimization of well locations and flow rates across real technological blocks, successfully demonstrating that dynamic flow redistribution—using methods like squared distance and streamline-based weighting—can reduce overall operational costs and reagent consumption by up to 5.7%.
However, resolving the complex systems of partial differential equations required for full 3D reactive transport modeling is notoriously computationally intensive. To drastically reduce computation times without sacrificing predictive accuracy, novel computational frameworks were introduced. The complex 3D simulation space was mathematically decomposed into sets of simpler 1D problems utilizing streamline and trajectory-based approaches. By coupling these streamlined methodologies with high-performance GPU parallelization and analytical solutions based on second-order approximation series, computational performance was massively accelerated. Validated against operational field data from deposits in Southern Kazakhstan, these optimized algorithmic methods achieved exceptional precision—yielding relative errors of less than 2%—providing a highly scalable, computationally inexpensive software solution for real-time well network management.
Artificial Neural Networks
To overcome the immense computational bottlenecks of traditional In Situ Leaching (ISL) simulation, which typically requires a sequential pipeline of geostatistical interpolation, computational fluid dynamics (CFD), and reactive transport modeling, I engineered a machine learning framework to forecast uranium extraction dynamics directly. By generating robust training datasets from both empirical experiments and complex 3D reactive transport simulations, I constructed and trained regression based artificial neural networks (ANNs) to predict mineral recovery rates over time. This data driven approach successfully bypasses the most computationally expensive modeling steps, delivering production extraction curves with a high degree of accuracy when compared to traditional, resource heavy mass transport models.
Building further on AI applications in geomodelling, Physics Informed Neural Networks (PINNs) were also explored to directly accelerate the underlying hydrodynamic simulations. Because solving the elliptic Poisson equation for hydraulic head consumes a substantial portion of computational resources during ISL scenario generation, PINNs and Inverse Distance Weighting (IDW) interpolation were utilized to calculate highly accurate initial approximations for the iterative solvers. By analyzing the impact of computational grids and well spacing constraints, this hybrid computational strategy was proven to drastically optimize the numerical analysis, cutting down the required number of solver iterations by a factor of 2.8 to 7.1 and significantly accelerating the overall simulation process.
Software Engineering
Over the past fifteen years, I have developed and deployed software solutions ranging from enterprise level industrial simulators to web portals. A primary focus of my work has been supervising the development of Simulator Version 3.0, a software modeling complex used to monitor mineral extraction via In Situ Leaching. Integrating this system directly into the Oracle production databases of JSC NAC Kazatomprom required bridging theoretical physics with live industrial operations. Additionally, I led the software development for eight research projects funded by the Ministry of Education and Science, translating complex scientific requirements into robust application architectures.
My technical foundation spans a versatile stack of programming languages and environments, including C++, C#, Python, Java, and modern web technologies like JavaScript, Node, Electron, and PHP. I actively build cross platform solutions, from accelerating 3D computational fluid dynamics simulations using Python and CUDA to developing modern Android applications. My standard engineering practice includes utilizing terminal based version control, configuring containerized Docker environments, and managing database migrations. This technical breadth builds on my earlier work developing GIS systems, reverse engineering educational platforms, and creating specialized design tools for oil refining reactors.
Beyond writing code, I integrate structured project management and certified IT expertise into the development lifecycle. I hold extensive official credentials, including a seven year tenure as a Microsoft Certified Trainer, a Cisco Certified Design Associate, and specialized Microsoft certifications in SQL Server database development and SharePoint administration. My formal training also includes international high performance computing schools, 3D geostatistics modeling in France, and comprehensive project risk and schedule management. This combination of system level engineering and architectural risk management ensures the consistent delivery of scalable, scientifically rigorous software tools.
Projects
Copyright certificates
Official recognition and intellectual property filings spanning algorithmic discoveries and proprietary codebase architectures.