Performance, stability, and control analysis of eVTOL vehicle. Integration with batteries, powertrain, control system, autonomy, structures, and aerodynamics. I combine surrogate modeling with domain specific knowledge to perform rapid and comprehensive analyses.

Preliminary design of eVTOL concepts. Intersection of conops, certification, and maneuverability performance. Control, optimization, and uncertainty quantification of robot+human manufacturing processes.

Developing DigitalTwin technology, combining operational data and physics-based lifing models to make more accurate lifing and maintenance decisions. Activities include data cleaning/processing, missing data estimation, usage modeling & forecasting, damage modeling (fracture mechanics), surrogate modeling, FEM, data fusion with Bayesian Networks, sensitivity analysis, inspection scheduling. I also work on advanced surrogate modeling techniques. I'm working to build a platform to enable multi-fidelity models, non-stationary and heteroscedastic processes, time- and space-dependent outputs, subspace methods, and take advantage of recent mathematical advances in MCMC methods (likelihood informed, subspace-adaptive). The goal is to bring state-of-the-art surrogate modeling to the industry.

Developing a method for adaptive unsupervised machine learning accelerated with adjoint information (physical/sample error estimates) from computational models.• Utilize cheap residual evaluations (separating physical errors and statistical model errors) to explore the high-dimensional parameter space• Optimize parameters of the statistical model• Detect well-modeled behavior to minimize samples• Implement an adaptive, 1D two-phase subcooled-boiling fluid model with adjoint solutions for testing the machine learning algorithmPerformed sensitivity and uncertainty study of various two-phase fluid flow models applied to a nuclear reactor core.• Compared Star-CD, Star-CCM+, and Nphase models• Used C, Fortran, and shell scripting to manage parallel runs on tens of cores• Uncovered 77% variation in void fraction due to bubble size model• Established lack of output convergence for commonly used meshes

Assissted development of a prototype hybrid engine battery charging system for an Au-tonomous Underwater Vehicle that extends mission time from 3 days to 1 month.• Redesigned electrical and sensor systems• Wrote control software for embedded microcontroller• Recharging system was successfully demonstrated for the US Navy

Extended an adaptive Cartesian-grid simulation suite (Chombo) to integrate CAD model geometries as embedded boundaries.• Implemented efficient computational geometry algorithms in C++• Enabled fluid simulations of complex nuclear reactor core geometry on a cluster of hundreds of processors

Developed, implemented, documented, and validated a new model for initial design and performance prediction of Vertical Axis Wind Turbines.• Incorporated blade interactions, dynamic vortices, a static stall model and various control strategies• Enabled design optimization with fast convergence• Achieved 3.4% efficiency gain with periodic pitch controlStudied plunging and pitching airfoils with in-house code and compared to experimental data.• Resolved earlier discrepancy due to insufficient mesh resolution• Validated 3D Discontinuous Galerkin method with moving boundaries

Measured reactive material debris clouds in pursuit of a model to predict post-break up characteristics, including particle size distribution. Obtained DoD clearance for work.

Evaluated software tools to help NASA systems engineers integrate new MBSE methods into the system design process.

Correlated Diffusion-Tensor MRI data to brain cancer recurrence patterns. Found evidence that treatment plans based on DT-MRI can be more effective and less damaging to healthy tissue.