Education

Ph.D. in Computational Science
Oden Institute for Computational Engineering and Sciences
The University of Texas at Austin, GPA 3.75 Expected May 2021 — Austin, TX

M.Sc. in Computational Science
Oden Institute for Computational Engineering and Sciences
The University of Texas at Austin, GPA 3.75 December 2018 — Austin, TX

B.Sc. in Computer Science (Hons.)
Cornell University, GPA 3.7 May 2016 — Ithaca, NY

Links

linkedin.com/in/sheroze1123
github.com/sheroze1123

Coursework

Graduate

Reinforcement Learning
Machine Learning
Bayesian Inference
Inverse Problems
Finite Element Methods
Numerical Analysis: Linear Algebra
Differential Equations
Functional Analysis
Mathematical Modeling

Undergraduate

Machine Learning
Computer Vision
Parallel Computing
Algorithms
Compilers
Operating Systems
Cloud Computing
Computational Physics

Publications and Projects

High-Dimensional Bayesian Inference — Python, TensorFlow, NumPy, PyMC3

April 2018 – Present
Oden Institute - The University of Texas at Austin

Accelerating high-dimensional parameter inference and uncertainty quantification incorporating dimensionality reduction and physics-informed neural network error models to efficiently perform Bayesian inference of simulation parameters with Markov Chain Monte Carlo using Hamiltonian dynamics. This work was presented at the 2019 SIAM Conference on Computational Science and Engineering at Spokane, WA and the 2019 DTRA annual conference.

Model-Aware Neural Networks for Compression — Python, TensorFlow, MPI, C++, NumPy

April 2018 – Present
Oden Institute - The University of Texas at Austin

Accelerating parameter inference by compressing time-series data of large-scale distributed physical models using model-aware neural networks trained in a data-parallel manner using Horovod on TACC supercomputers Maverick2 and Lonestar5.

Regularized Parameter Inference — Python, NumPy, SciPy, FEniCS, hIPPYlib

August 2016 – April 2018 Oden Institute - The University of Texas at Austin

Implemented a limited-memory higher order optimization routine to infer parameters in partial differential equations and quantified uncertainty using Markov Chain Monte Carlo methods.

Professional Experience

Facebook Reality Labs, Redmond - Research Intern

Extended a block automatic differentiation framework to accelerate nonlinear optimization problems on differentiable manifolds to calibrate sensors from observations. Implemented a modular simulation framework to enable uncertainty quantification and faster development of unit tests and benchmarks.

Goldman Sachs, NYC — Technology Analyst

Extended a Sybase IQ log parser to capture executed statements and associated metadata. Created Kibana dashboards to display IQ server diagnostics. Wrote an efficient bulk renderer to offload data onto an Elasticsearch instance.

Teaching — Numerical Linear Algebra, Computer Vision, Operating Systems

Prepared student assignments including coding a microkernel in C, automated panorama stitching, convolutional networks, randomized singular value decomposition methods and algebraic eigenvalue problems.

Academic Distinctions

Graduate Fellowship — Oden Institute - The University of Texas at Austin

Awarded the National Initiative for Modeling and Simulation Fellowship.

Gene Golub Summer School — Breckenridge, CO

Accepted to the 2018 Gene Golub SIAM Summer School on Inverse Problems: Systematic Integration of Data with Models under Uncertainty.

Talks

Mitigating the Cost of PDE-constrained Bayesian Inverse Problems Using Dimensionality Reduction and Machine Learning - 2019 SIAM Conference on Computational Science and Engineering

Bayesian Inverse Problems Using Dimensionality Reduction and Machine Learning - 2nd Annual Meeting of SIAM Texas-Louisiana Section 2019

Bayesian Inverse Problems Using Dimensionality Reduction and Machine Learning - 2020 SIAM Conference on Uncertainty Quantification

Skills

Technical

Machine Learning: regression, dimensionality reduction, unsupervised learning, convolutional autoencoders, recurrent neural networks, Bayesian neural networks.
Statistics: Bayesian inference, MCMC, uncertainty quantification, model validation.
Scientific Computing: quasi-Newton optimization methods, regularized inverse problems, numerical differential equations, Bayesian optimization, supercomputing.

Programming

Python (NumPy, SciPy, TensorFlow, scikit-learn, Keras, Pandas, Matplotlib, PyMC3)
C/C++, Java, MATLAB, MPI, OpenMP, FEniCS, PETSc/Tao, UNIX, CMake, Git