Gajanan Choudhary

E-mail: [email protected]

Address: Austin, TX, USA

Phone: +1 (512) 657-3030

LinkedIn: https://www.linkedin.com/in/gajananchoudhary

GitHub: https://github.com/gajanan-choudhary

Website: https://gajanan-choudhary.github.io

Summary

Computational scientist with 12+ years of interdisciplinary research and high-performance computing (HPC) production software development experience through long-term projects spanning applied mathematics, CPU/GPU kernel development, optimization, and machine learning. Creator of 4 scientific software, contributor to 5 HPC software written in C/C++, SYCL, Python, and Fortran, maintainer of the sparse BLAS domain in the open-source oneAPI Specification under the Unified Acceleration (UXL) foundation, and author of 6 technical publications.

Skills

• Software development:

  • Programming: C/C++, SYCL/DPC++, Python, Fortran, MATLAB, MPI, OpenMP, f2py, SWIG, Python/C API, and Bash.
  • Tools: Git, GitHub, Copilot, Bitbucket, Mercurial, SVN, Travis CI, CircleCI, Docker, Coveralls, Codecov, CMake, GNU Make, Gcov, LCOV, GProf, GDB, Valgrind, Doxygen, LaTeX, HTML, and CSS.
  • Concepts: High-performance computing, parallel programming, CPU and GPU performance optimization, kernel development, x86 intrinsics, assembly, neural networks, deep learning, large language models, transformers, library engineering, language interoperability, data structures, algorithms, complexity, object-oriented programming (OOP), standard template library (STL), debugging, code coverage, continuous integration and continuous delivery (CI/CD), and test-driven development (TDD).

• Research:

  • Engineering: Computational mechanics, computational fluid dynamics (CFD), solid mechanics, structural dynamics, fluid-structure interaction (FSI), and coupled models.
  • Mathematics: Linear algebra, functional analysis, partial differential equations, numerical methods, advanced theory of finite element methods, approximation, and optimization.
  • Applications: Adaptive Hydraulics (AdH), ADvanced CIRCulation (ADCIRC), Gridded Surface Subsurface Hydrologic Analysis (GSSHA), Aquaveo SMS/WMS, ANSYS, Abaqus, ParaView, AutoCAD, and STAAD Pro.

Work Experience

• Math Algorithm Engineer / Software Engineer, Intel Corporation, Austin, TX, February 2021 - Present.

  • Responsibilities: Owned sparse linear algebra and high-performance conjugate gradient (HPCG) benchmark components as a contributor in the Intel® oneAPI Math Kernel Library (oneMKL) team.
  • Led the redefinition of sparse BLAS domain APIs in the oneAPI Specification for the UXL Foundation as a maintainer, and directed and reviewed implementation of Intel oneMKL, NVIDIA cuSPARSE, and AMD ROCm backends in the foundation's hardware- and platform-agnostic open-source implementation of those APIs: the oneMath library.
  • Implemented GPU performance optimizations for the sparse matrix-vector product (GEMV) and sparse × sparse matrix product (spGEMM) SYCL APIs for compressed sparse row (CSR) matrices.
  • Received division awards for implementing 700+ high-quality, general, hand-optimized wrappers over x86 intrinsics APIs to support SIMD lengths {1, 2, 4, 8, 16} on AVX2 and AVX512 architectures for all common integer and real/complex floating-point types to ease kernel development, performance optimization, and testing overhead by orders of magnitude.
  • Implemented extreme CPU performance optimizations for CSR and block compressed sparse row (BSR) matrix formats for sparse × dense matrix multiplication (GEMM) API in oneMKL for AVX2 and AVX512 architectures for all supported types using vectorization, unrolling, x86 intrinsics wrappers, and assembly.
  • Introduced optimized SYCL/GPU APIs in oneMKL for sparse matrix addition, sorting, and transposition.

• Research Associate, The University of Texas at Austin, Austin, TX, December 2020 - February 2021.

  • Responsibilities: Conducted research on hurricanes and managed research direction of post-doctoral fellows and doctoral students in the team.
  • Led the development of coupled PyTorch neural network and CFD/physics models in the Python software, Water Coupler, to potentially save lives and billions of dollars through hurricane compound flood forecasts.
  • Led software projects that the US Army Corps of Engineers (USACE) sponsored to expand the capabilities of their CFD software, AdH, through addition of PETSc solvers and coupling with HEC-RAS.

• Postdoctoral Fellow, The University of Texas at Austin, Austin, TX, October 2019 - December 2020.

  • Responsibilities: Open-source parallel software development for research on coupled fluid dynamics models.
  • Coupled diffusive wave and groundwater models in the C software, Adaptive Hydraulics (AdH).
  • Created a Python HPC interface for the Fortran software, ADvanced CIRCulation (ADCIRC).
  • Coupled the C++ software, Gridded Surface Subsurface Hydrologic Analysis (GSSHA), with ADCIRC in the Python software, Water Coupler, for simulation of compound floods from hurricanes.

• Graduate Research Assistant, The University of Texas at Austin, Austin, TX, September 2014 - October 2019.

  • Responsibilities: Software and library development for research on coupled fluid dynamics models.
  • Created, designed, implemented, tested, and validated the Python software, Water Coupler, for coupling hydrodynamic and hydrologic models written in C, C++, Fortran, and Python.
  • Created, the Python software, htopy, for partially automating the generation of platform-independent Python wrappers of software written in C and C++, and used it to create parallel Python interfaces of AdH (C) and GSSHA (C++) in minutes instead of days.
  • Coupled 2D and 3D shallow water models in AdH for enabling baroclinic 3D models with wetting and drying, a capability that almost none of the existing 3D models have.
  • Made numerous contributions to AdH, including developing its wind library, adding mesh adaption for prism elements, and improving parallel load rebalancing of 3D models.
  • Experience working on federal projects in collaboration with the Engineer Research and Development Center, US Army Corps of Engineers, funded by the Department of Defense under the High Performance Computing Modernization Program.

• Assistant Surveyor, Indian Register of Shipping, Mumbai, India, July 2013 - July 2014.

  • Responsibilities: Research on stress response of ship hulls to bending, shear, torsion, and warping loads.
  • Created, tested, verified, validated, and documented a MATLAB software with a graphical user interface (GUI), IR-SECT, for 2D modeling of ships and for calculation of shear flow and sectorial properties of ship cross-sections using graph algorithms.
  • Worked on 3D modeling and finite element analysis of a post-Panamax container ship using ANSYS.
  • Work on sectorial properties and torsion published externally as a company guideline in 2024.

• Structural Design Intern, Himanshu Tulpule and Associates, Pune, India, May 2012 - July 2012.

  • Responsibilities: Structural design and detailing of reinforced concrete and steel structures.
  • Designed 14 structures spanning numerous types, including trusses, beams, columns, slabs, foundations, arches, and joints.
  • Conducted site visits for studying topography to help with planning and design.

Education

• The University of Texas at Austin, Austin, TX, August 2014 - December 2019, Doctorate (Ph.D.) in Engineering Mechanics, GPA: 3.92.

• The University of Texas at Austin, Austin, TX, August 2014 - May 2018, Graduate Portfolio in Scientific Computing, GPA: 4.00.

• The University of Texas at Austin, Austin, TX, August 2014 - May 2017, Master of Science (M.S.) in Engineering Mechanics, GPA: 3.89.

• Indian Institute of Technology (IIT) Kharagpur, Kharagpur, India, July 2009 - July 2013, Bachelor of Technology (B.Tech.) in Civil Engineering, GPA: 8.23/10.

Software contributions

• Maintainer, Sparse BLAS component in the oneAPI Specification: Designing and maintaining portable, performant, platform-agnostic, and hardware-agnostic sparse BLAS SYCL math APIs.
• Contributor, Intel® oneAPI Math Kernel Library (oneMKL): Owned feature implementations and CPU/GPU architecture-specific performance optimizations for sparse linear algebra and HPCG benchmark components in oneMKL, written in C, C++, SYCL/DPC++, and Fortran.
• Creator, Water Coupler: Developed the open-source parallel Python software for coupling physics models, AdH (C), GSSHA (C++), and ADCIRC (Fortran), as well as PyTorch machine learning/neural network models for real-time hurricane compound flood simulations.
• Contributor, Adaptive Hydraulics, US Army Corps of Engineers: Created the C/Python interface of AdH (adhpython), enhanced its meteorological library, improved mesh adaption, and coupled its shallow water, ground water, and diffusive wave models, all in parallel.
• Contributor, ADCIRC: Created the Fortran/Python interface of ADCIRC (pyADCIRC), enabling physics-based machine learning applications and multi-software coupling in Water Coupler.
• Contributor, GSSHA: Created the C++/Python interface of GSSHA (gsshapython) enabling multi-software coupling with ADCIRC and AdH in Water Coupler.
• Creator, htopy: Developed an open-source Python software for partially automating the generation of Python interfaces of C/C++ software, and used it to generate the Python interfaces of AdH and GSSHA.
• Creator, AdH-meshbuilder: Developed an open-source C software for generating arrays of coupled AdH finite element models.
• Creator, IR-Sect: Developed a proprietary MATLAB software with a graphical user interface (GUI) for modeling 2D ship cross-sections and calculating their sectorial properties by implementing graph algorithms.

Projects

• The University of Texas at Austin, Austin, TX; Supervisor: Dr. Clint Dawson.

  • Coupling ADCIRC and GSSHA for actionable forecasts of compound floods -- October 2019 to Present.
  • PyADCIRC: A Python interface for ADCIRC for multi-software coupling -- October 2019 to Present.
  • Coupling 2D overland and 3D groundwater flow models in AdH -- October 2019 to Present.
  • Coupling AdH, GSSHA, and NAM for actionable forecasts of compound floods -- August 2018 to October 2019.
  • Algebraic coupling of 2D and 3D shallow water and transport FE models -- September 2015 to October 2019.
  • Adding mesh adaption capability for 3D prism elements in AdH -- May 2018 to August 2018.
  • Improving parallel load rebalancing of adaptive tetrahedral meshes in AdH -- September 2017 to May 2018.
  • Numerical solution of multivariate first order ordinary differential equations -- October 2017 to December 2017.
  • Adding parallel XDMF output support to AdH for improved I/O and visualization -- May 2017 to August 2017.
  • Parallel meteorological library development for adaptive meshes in AdH -- April 2015 to August 2015.
  • Hurricane vulnerability of the Houston-Galveston Area Protection System -- September 2014 to March 2015.

• Indian Register of Shipping, Mumbai, India; Supervisors: A. R. Kar and K. M. Doshi.

  • FE analysis of a post-Panamax container ship using ANSYS -- January 2014 to July 2014.
  • Sectorial properties of thin-walled beams with arbitrarily shaped cross-sections -- November 2013 to July 2014.
  • Automatic calculation of shear flow in ship hull girders -- July 2013 to December 2013.

• Indian Institute of Technology (IIT) Kharagpur, Kharagpur, India; Supervisors: Dr. N. Mitra and Dr. S. K. Barai.

  • Near-field blast response of sandwich composites submerged under water -- February 2013 to June 2013.
  • Discrete optimization of truss weight using genetic algorithm -- August 2012 to May 2013.
  • A MATLAB software with a GUI for graphical modeling and FE analysis of trusses -- August 2012 to December 2012.
  • Crack detection in concrete surfaces using image processing, fuzzy logic, and neural networks -- March 2012 to October 2012.

Activities

• Diversity: Mentored an undergraduate student in Spring 2018 as part of Graduates Linked with Undergraduates in Engineering (GLUE) program at UT Austin, aimed at inspiring women to pursue graduate degrees in STEM.

• Dance: Completed intermediate level training in various social dance forms at UT Austin and Austin Swing Syndicate.

• Teaching: Attended workshops on teaching and obtained a basic teaching certificate from UT Austin in Fall 2019.

• Music: Received formal training in classical piano at UT Austin and performed as a drummer in competitions at IIT Kharagpur.

Peer-reviewed Publications

• Choudhary, G. K., Trahan, C. J., Pettey, L., Farthing, M., Berger, C., Savant, G., Inanc, E., Dawson, C., & Loveland, M. (2025). Strongly coupled 2D and 3D shallow water models: Theory and verification. Journal of Hydraulic Engineering, 151 (1), 04024049.

• Trahan, C. J., Savant, G., Berger, R. C., McAlpin, T. O., Pettey, L., Choudhary, G. K., & Dawson, C. N. (2018). Formulation and application of the Adaptive Hydraulics three-dimensional shallow water and transport models. Journal of Computational Physics, 374, 47-90.

• Choudhary, G. K., & Doshi, K. M. (2015). An algorithm for shear stress evaluation in ship hull girders. Ocean Engineering, 108, 678-691.

• Choudhary, G. K., & Dey, S. (2012, October). Crack detection in concrete surfaces using image processing, fuzzy logic, and neural networks. In Advanced Computational Intelligence (ICACI), 2012 IEEE Fifth International Conference on (pp. 404-411). IEEE.

Non-peer-reviewed Publications

• Choudhary, G.K. (2019). Coupled atmospheric, hydrodynamic, and hydrologic models for simulation of complex phenomena (Doctoral Dissertation). The University of Texas at Austin, USA. Retrieved from https://dx.doi.org/10.26153/tsw/7729.

• Choudhary, G.K. (2017). Algebraic coupling of 2D and 3D shallow water finite element models (Master’s report). The University of Texas at Austin, USA. Retrieved from https://doi.org/10.15781/T2Q81570X.

Technical Reports

• Choudhary, G. K., & Doshi, K. M. (2024). Guidelines on assessment of hull girder stresses due to torsion in ships with large deck openings. Indian Register of Shipping, India. Retrieved from https://www.irclass.org/media/7468/guidelines-torsion_irs-g-des-17_sept-2024_new.pdf.

• Choudhary, G. K., & Dawson, C. (2014-2019). Over 20 reports submitted to the Engineer Research and Development Center and the Department of Defense for 4 projects sponsored under PETTT HPCMP program of DoD.

• Choudhary, G. K. (2018). Verification and validation of coupled 2D and 3D shallow water finite element models (Unpublished report for Graduate Portfolio in Scientific Computation). The University of Texas at Austin, USA. Available upon request.

• Numerous authors. (2015). SSPEED Center Report on Houston-Galveston Area Protection System (Annual report). Rice University, USA. Retrieved from http://speed.rice.edu/sspeed/downloads/HGAPS_Report_08_31_15.pdf.

• Choudhary, G. K., & Doshi, K. M. (2014). An algorithm for evaluation of sectorial properties of arbitrarily shaped prismatic thin-walled beams (Internal publication). Indian Register of Shipping, India.

• Choudhary, G. K. (2013). Discrete optimization of truss weight using genetic algorithm (Unpublished bachelor’s thesis). Indian Institute of Technology Kharagpur, India. Available upon request.

Conference Talks

• Choudhary, G. K., & Dawson, C., 2020, PyADCIRC: A Python interface for accessing functions and variables of ADCIRC in Python, 2020 Virtual ADCIRC Users Group Meeting, Mar 30-31.

• Choudhary, G. K., Trahan, C., Pettey, L., Farthing, M., & Dawson, C., 2017, Algebraic coupling of 2D and 3D shallow water finite element models, The 16th International workshop on Multiscale (Un)-structured mesh numerical Modeling for coastal, shelf, and global ocean dynamics, Stanford, CA, USA, Aug 29-Sep 1.

• Choudhary, G. K., & Dey, S., 2012, Crack Detection in concrete surfaces using image processing, fuzzy logic, and neural networks, 2012 IEEE Fifth International Conference on Advanced Computational Intelligence (ICACI), Nanjing, China, Oct 18-20.

Languages

• English: Fluent

• Hindi: Native

• Marathi: Native