Summary
Dinesh Kumar is a research scientist based in Berkeley with 11 years of experience applying machine learning, high-performance computing, and computational mathematics to imaging and inverse problems. He has built GPU-accelerated toolchains (C/C++, CUDA, Python) for non-uniform FFTs and massive parallel simulation suites used in X-ray scattering and tomographic reconstruction. His work spans numerical methods from mesh-free geoflow simulators to finite-element nonrigid registration, demonstrating a rare combination of rigorous PDE/numerical modeling and production-grade HPC engineering. At Berkeley Lab he moved from postdoc to project scientist and now research scientist, leading practical pipelines for calibration, feature extraction, and reverse Monte Carlo reconstruction. Colleagues rely on him to translate complex physics models into efficient, open-friendly software that integrates with community packages like TomoPy. He often pairs deep domain knowledge with hands-on optimization on NVIDIA GPUs, enabling orders-of-magnitude speedups for large-scale scientific imaging.
11 years of coding experience
11 years of employment as a software developer
BE, Mechanical Engineering, BE, Mechanical Engineering at Regional Engineering College, Silchar
PhD, Mechanical Engineering, PhD, Mechanical Engineering at University at Buffalo
MS, Mechanical Engineering, MS, Mechanical Engineering at State University of New York at Buffalo