I build machine learning models for engineering and scientific systems, and I test whether they can be trusted.

MS in Aerospace Engineering from Georgia Tech (Aerospace Systems Design Laboratory). Previously eighteen months on airworthiness certification for the Rolls-Royce Trent XWB-84 EP, where a model being wrong was not an academic problem.

Most of my work returns to the same question: when a model produces a number, what would have to be true for that number to mean anything?

Model validation — leakage control, held-out protocols matched to data structure, and tests designed to fail when the pipeline is wrong.

Uncertainty quantification — Gaussian processes, surrogate error characterisation, parametric uncertainty propagation.

Surrogate & reduced-order modeling — multifidelity datasets, proper orthogonal decomposition, Operator Inference.

Data quality at scale — finding the failures that pass every standard sanity check.

Equity-Backtest — A monthly cross-sectional equity strategy built to a strict walk-forward protocol, published with its own negative result. Net Sharpe 0.25 (t = 0.81), rising only to 1.07 with costs set to zero. Every specification tested is logged and committed, including the one that looked best.

NURBS_BEM_EMSolver — A mesh-free boundary element solver for solenoid magnetic fields on NURBS geometry, built from mathematical formulation through code, with parametric multifidelity datasets for surrogate modeling and Operator Inference.

Surrogate-model-learning — Gaussian process, response surface and RBF surrogates compared across benchmark problems, focused on where each estimator breaks down rather than which one wins.

atlanta-mobility-resilience-digital-twin — Road-network model of Atlanta from OpenStreetMap: disruption scenarios, origin–destination travel-time comparison, accessibility under stress.

Python · pandas · NumPy · SciPy · scikit-learn · PyTorch · NetworkX · Git · pytest · MATLAB · C++

Portfolio · LinkedIn · tsarkar34@gatech.edu

How to use this file: create a new public repository named exactly triasha72 (same as your username), add this as README.md, and GitHub renders it at the top of your profile page. It's the first thing a recruiter sees when they click your GitHub link from a CV.