Leveraging LLM/NLP on the next generation of smart glasses

Built DL models for an auto insurance claim to be processed with zero human intervention

Developed a predictive model on repair cost prediction services which has been adopted by several major insurance companies in the US

Delivered a vehicle cluster model by their potential liability and medical payments, which can be used for insurance premium estimation directly, or as input features for other predictive models

- Developed a support vector machine (SVM) model for credit scoring with over 30,000 observations by SAS- Solved the imbalanced data problem by using least square weighted SVM (LS-WSVM) approach- Implemented suitable recursive feature elimination (RFE) algorithm for variable selection- Achieved a lift of 0.93% for in-time validation and 3.42% for out-of-time validation in KS statistics compared to the previous logistic model used in production, which could bring more than 1 million dollars in revenue

Teaching calculus, linear algebra and differential equation classes for undergraduate students

Performing numerical analysis on nonlinear partial differential equations- Developed discrete numerical solver for nonlinear Schrodinger partial differential equations in MATLAB- Categorized and examined solutions of discrete solitons and vortices with respect to the persistence and stability of the PDE solutions by means of the Lyapunov-Schmidt theory- Evaluated the performance of the numerical solver by analyzing the deviation from the analytical solution- Identified the bifurcation diagrams of stationary solutions and their spontaneous symmetry breaking points- Simulated the dynamical behavior of the stationary solutions as well as the ghost solutions where they exist- Published 6 research papers, 3 of them are as first author