Main Accomplishments

Significant experience in developing numerical codes for a broad array of physics/engineering problems and deep learning (DL) / uncertainty quantification (UQ) tools

• Development of a computationally efficient Python framework to predict $\mathrm{CO}_2$-injection induced seismicity using multifidelity Monte Carlo and graph neural networks (GNNs).
• Development of a finite-element based method to compute the damping rate of fluid oscillations in microfluidic nozzles that is orders of magnitude faster than computational fluid dynamics (CFD) approaches.
• Contribution to design and testing of a novel computational model for simulation of compressible flows involving shocks and contact discontinuities using a localized artificial diffusivity (LAD) approach.
• Use of operator learning to predict initial-to-final state mapping of droplet deposition in a moving subdomain approach for generating 3D printed parts in order to accelerate part-level quality prediction.
• Construction of surrogate models based on physics-informed machine learning (ML) and deep neural networks for accelerating the discovery of novel materials and the design of energy storage devices such as Li-Ion batteries and supercapacitors.
• Design and implementation of novel Monte Carlo-based solvers to accelerate predictive modeling/UQ for multiphase flows in subsurface phenomena.
• Development of numerical solvers for a wide range of nonlinear systems with inherent uncertainty or random forcing.
• Construction of innovative models for shocked particle-laden flows that enable process-scale, predictive simulations of high-speed applications ranging from dust explosions to engines for hypersonic propulsion systems.

Current Employment

ENERGY SCIENCE & ENGINEERING DEPARTMENT - Stanford, CA, USA

• Position: Physical Science Research Scientist, September 2024-Present
• Duties:
  • Development of a multifidelity Monte Carlo framework for fast, accurate estimation of cumulative distribution functions of quantities of interest in induced seismicity.
  • Integration of fast surrogate models based on graph neural networks into this multifidelity approach.
  • Guidance of current PhD students working on multilevel/multifidelity approaches in multiphysics problems.

Past Employment

FUTURE CONCEPTS DIVISION, SRI INTERNATIONAL - Palo Alto, CA, USA

• Position: Research Scientist, May 2023-June 2024

PARC, A XEROX COMPANY - Palo Alto, CA, USA

• Position: Member of Research Staff, October 2020-April 2023

DEPARTMENT OF ENERGY RESOURCES ENGINEERING, STANFORD UNIVERSITY - Stanford, CA, USA

• Position: Postdoctoral Scholar, September 2018-September 2020
• Duties:
  • Development of enhanced multilevel Monte Carlo methods for accelerating the estimation of cumulative distribution functions of quantities of interest in multiphase flows e.g. in stochastic reservoir simulation:
    • Fusion of standard Monte Carlo with variance reduction techniques to speed up UQ.
    • Sponsored by Air Force Office of Scientific Research (AFOSR) and Total S.A. and aimed at inventing, testing and deploying new UQ techniques that are more computationally efficient and yet as accurate as current exhaustive system realizations.
  • Design of a DNN with TensorFlow 2 to accelerate sensitivity analysis and UQ of energy storage systems to enable rapid design prototyping.

SAN DIEGO STATE UNIVERSITY RESEARCH FOUNDATION - San Diego, CA, USA

• Position: Research Specialist I, October 2016-July 2018
• Duties:
  • Development of transformative Eulerian-Lagrangian methods for simulation of particle-laden flows in high-speed engineering applications:
    • Design and validation of Cloud-In-Cell models with enhanced accuracy and efficiency compared to the current state of the art for simulation of a particle cloud interacting with a shocked carrier flow.
    • Multi-institution collaborative effort aimed at designing a general multiscale framework for modeling multimaterial dynamics.
    • Sponsored by Air Force Office of Scientific Research (AFOSR) and aimed at designing and implementing computational models capable of predicting the dynamics of process-scale, compressible particle-laden flow problems with uncertain parameters.

THEORETICAL DIVISION, LOS ALAMOS NATIONAL LABORATORY (LANL) - Los Alamos, NM, USA

• Position: Graduate Research Assistant, August 2014 to June 2015
• Duties:
  • Development of an information-directed approach for materials discovery and design:
    • Lab-Directed Research and Development (LDRD) effort (>$1M) aimed at demonstrating the capability to accelerate materials discovery.
    • Design of a physics-informed ML technique for model selection in materials science and beyond.

Skills

• Computing:
  • Programming languages: C/C++ (basic), Fortran (basic), Python (intermediate) including NumPy and SciPy, MATLAB (intermediate)
  • Unix/Linux, basic HPC, AWS, GitHub
  • Windows, Mac OS and Linux
• Numerical and Statistical Methods/Software Packages:
  • Finite difference/volume
  • Higher-order methods
  • Particle-in-cell, Eulerian-Lagrangian
  • Compressible flows, shock physics
  • Subsurface flows
  • Statistical mechanics
  • Monte Carlo simulation, stochastic collocation
  • Applied probability and Uncertainty quantification
  • Software packages: COMSOL (multiphysics), OpenFOAM (multiphysics), FEniCS (finite element), Mathematica
• Deep Learning:
  • Software: PyTorch (preferred), Tensorflow 2 (basic)
  • Operator learning (Fourier Neural Operator, DeepONet)
  • Basics of standard neural networks (CNN, RNN, etc.) via self-study of online courses (Stanford CS231n (Convolutional NNs) and CS221 (Artifical Intelligence), Coursera (Deep Learning Specialization))
  • Basics of graph neural networks (GNNs) via self-study of online courses (Stanford CS224W) and project work ($\mathrm{CO}_2$-injection induced seismicity)
  • Physics-informed machine learning (e.g., PINN)
• Editing/Presentation:
  • Sublime Text, LaTex (including Beamer), Markdown
  • Word, PowerPoint, Keynote
• Written/Oral Communication and Leadership:
  • Preparation and peer review of scientific manuscripts
  • Interdisciplinary teamwork and student/intern research supervision both in industry and academia
  • Project presentations at international scientific conferences