I developed an innovative methodology for automating the creation of neural architectures using large language models (LLMs) and nature-inspired algorithms. This ongoing project, implemented in Python and deployed on Google Cloud, leverages both closed and open-source LLMs alongside various nature-inspired algorithms.

Under Dr. Iacob's mentorship, I spearheaded a comprehensive research initiative centered on developing and implementing advanced machine learning techniques. My work primarily involved autoencoding and other self-supervised learning methods, aiming to accurately classify large sets of unlabeled image data. Our efforts culminated in achieving a high level of classification accuracy.

I researched and implemented various neural operators, focusing on their applications to Navier-Stokes and Darcy Flow partial differential equations (PDEs). These neural operators were trained and tested on mesh-based data, aiming to match the accuracy of traditional mesh methods while significantly increasing inference speed.

I developed custom Gymnasium environments for various forms of the knapsack problem and implemented several deep reinforcement learning (DRL) algorithms to accelerate computations. Our solutions were highly accurate and state of the art in their domain.

I was the full stack developer completing all steps of the Software Development Lifecycle in this project.I elicited requirements from the client, made documents, developed, tested and deployed a dynamic and configurable survey hosting website for the Department of Education at Georgia Southern University(GSU). I implemented security using Microsoft Identity Manager and rights based access. I additionally anonymized all data collected ensuring the privacy of survey participants. The software consisted of a Microsoft SQL Server, Asp.Net backend and a Blazor client front end.

In this project I implemented a variety of models in order to analyze and predict data. These models include creating a fuzzy system,hidden markov models, and xgboost classification models. I then used these models to analyze aspects of the data I was provided to find points of interest and predict metrics for the Professor I was working for.

Used deep reinforcement learning (DRL) to approximately solve np-complete operations research problems. The process involved creating Gymnasium environments along with implementing and evaluating various DRL algorithms on metrics such as accuracy and inference speed