Who i am

About Me

Math PhD from UC Berkeley seeking an AI / ML role

Previously a software engineer at Google on the Trust & Safety team

I have over 7+ years of experience in software engineering and mathematical research. My core skills include machine learning, numerical analysis, scientific computing, and teaching. I am passionate about applying my skills to solve real-world problems and make a positive impact on society. On the weekends, I love to swim, bike and run.

ML/Data Science Projects

2024

GITHUB PROJECT #1

Classify MNIST data with Tensorflow/Keras

Methods: LDA, neural nets

2024

GITHUB PROJECT #2

Analyze men’s D1 NCAA times in the 500 freestyle

Cluster swimmers based on race strategy

Predict final times from the first few splits

Methods: Ridge and LASSO regression

research

Publications

2023

Numerische Mathematik

2023

SIAM Journal on Matrix Analysis and Applications

Description: Analyze floating-point summation error for large/low-precision problems on arbitrary summation trees.

Description: Proposes a method for finding a low-rank approximation to f(A) using Krylov subspace projections. Gives applications to trace estimation.

research

Publications

2023

SIAM Journal on Matrix Analysis and Applications

Description: Provides accuracy bounds on Monte Carlo estimators for the diagonal of a symmetric matrix.

2023

SIAM Journal on Scientific Computing

Description: Proposes randomized algorithms for compressing tensors stored in tensor-train format; compares with deterministic methods.

research

Publications

2022

SIAM Journal on Matrix Analysis and Applications

Description: Block Lanczos matrix sketching is competitive with randomized subspace iteration.

2022

Linear Algebra and its Applications

Description: Uses multilevel Monte Carlo techniques to approximate the trace of f(A) by splitting the approximating polynomial into several smaller ones.

research

Publications

2021

Applied Mathematics Letters

2020

Linear Algebra and its Applications

Description: Estimates the trace of a matrix polynomial 2x faster compared to standard methods.

Description: New backward error estimate for AX=B (multiple columns) guaranteed within 2x of true value.

research

Publications

2020

SIAM Journal on Matrix Analysis and Applications

2018

SIAM Journal on Matrix Analysis and Applications

Description: Proposes error estimate when solving Ax~=b; under certain assumptions it can’t be improved.

Description: Proposes new method LSMB for solving Ax~=b, offers efficient backward error estimation. Iterates are linear interpolation of LSQR/LSMR.