Dmitry Firsov Email & Phone Number

Santa Clara, California, Us

Research and develop in C++11; document numerical algorithms for Computational Electromagnetics (CEM) software to improve EM simulator that can be launched on 300GB memory workstation with GPUCarry out code refactoring, performance/memory optimization and documentation of numerical algorithmsTook role of team lead to dramatically increase accuracy, stability, and performance of Lorentz’s two main computational EM engines, with performance of most important direct solver engine improved at least 4x timesEnsure release note for each new version contains speed, memory usage, or accuracy improvementsDiscovered how to apply Helmholtz decomposition using algebraic analysis to improve EM formulation; decreased condition number of final system of linear equations to use single accuracy precision in direct solver: gained 2x performance boost, halved memory, solved problems on regular GPUs without fast double precisionOrganized total rebuild of frequency domain integral equations iterative EM solver (previously failed project) based on ACA compression using GMRES type algorithm; helped bring iterative solver from prototype to production-level with same accuracy as direct engine on much smaller memoryTo improve EM Integral Equations Solver, developed Krylov subspace linear algebra numerical algorithms, made precondition matrix optimizations, round-off error optimization, can solve very large systems of linear equations with ACA compressed matrices, C++ OpenMP, multi-threading/thread pools, SIMD, MKL, GPU MAGMA parallelizationLeveraged advanced math expertise to help rebuild and accelerate iterative solver using round-off errors treatment; improved condition number and appropriate normalization of system of linear equationsContribute to performance/accuracy of EM engine, directly increasing likelihood of long-term support contractsOptimized existing Numerical Linear Algebra algorithms; improved matrices condition number and EM models

Calgary, Alberta, Ca

Researched and developed in C++ geological cross-platform Oil&Gas CAD - Data Analysis Tool, 3D geological geometry, grid generation/ manipulation, partial differential equations solvers, statistics/stochastics algorithms, inverse problems, UI development, optimization algorithms, Numerical Linear Algebra, Image Processing/3D Seismic data processingDeveloped advanced mesh generator for detailed stochastic 3D modeling of reservoirs and underground structures such as channels, barforms, splays, rivers deltas, lobes.Developed numerical algorithms for large set of geological/geophysical simulations and data analysis toolsDeveloped parallel mesh to mesh interpolation to enable mapping of properties between totally independent structured and unstructured gridsHandled large problems associated with PDE-based 3D seismic data denoising tools: generalized to use same nonlinear implicit PDE solver with large sparse matrix; solver reused in elliptic grid generator for modeling complex geometries

Winnipeg, Mb, Ca

Primary developer of 3D Time Domain Maxwell Partial Differential Equations solver featuring 3D Finite Volume Time Domain Method coupled with Integral Equation, unstructured tetrahedral mesh to handle design of real objects, and supported dielectrics, perfect electrical conductors, lossy materials, thin wires, resistors, capacitors, and voltage sources to 2nd and 3rd order of accuracyKey member in group of professors and PhD students; co-supervised 2 undergraduates studentsMade significant contribution to EM research; team paper received 2007 ACES Outstanding Paper awardC++ with OpenMP and MPI, cross-platform, GUI based on wxWidgets with basic post-processing, "c-style" scripting for project files, database of materials, data export to VTK

Winnipeg, Mb, Ca

Developed numerical algorithms for image processing with C++ with OpenMP and MPIHelped create fast and efficient deblurring algorithm and efficient parallelization of image denoising algorithmsAnalysed math theory behind parallel simulations

Tomsk, Russia, Ru

Supervised undergraduate and graduate students; co-supervised 1 PhD and supervised 5 Master-level projects Taught Theoretical Mechanics, Scientific C++ Programming, and Numerical MethodsDesigned C++ implicit solver for Navier-Stokes equations to model 3D fluid flow in complex domain and simplified model of conductive liquid flow under magnetic field influence: extensive simulations required solution of huge sparse systems of linear equations by Krylov subspace methods with advanced precondition matrices; own implementation of Krylov subspace algorithms accelerated with round-off errors treatmentFinancial mathematics partial differential equations solvers

Doctor Of Philosophy (Ph.D.), Computational Fluid Dynamics

Master'S Degree, Mathematics