Within my time at UF, I have worked on a handful of projects that center around data science, artificial intelligence, and ethics. While my projects all exercise a different skillset, they focus on technology that is beneficial and positive. On one project, MoDA, I assist with the creation, implementation, and testing of an in-store Android mobile shopping assistant that advises the user to the products that most meets their requests. In another project, I am collecting sentiment on recent advances in Artificial Intelligence (AI) to determine if perspectives of AI and the social impact of AI differ across socio-economic, racial, gender, and/or geographical line. In my final project, I test the impact user domain knowledge has on AI explainability by measuring the trust individuals have with a simulated machine learning tool that generates mostly incorrect responses with believable explanations.

Utilized novel techniques from robust statistics to create a new methodology for adding a dimension of explainability to black-box models.Tested novel technique on a diverse selection of use cases including datasets and model types.

Collaborated across the company as an algorithmic bias consultant, assisting teams with fairness concerns in their differing applications of machine learning & Exposed several teams and employees to new and emerging fairness AI technologies and methods for addressing algorithmic bias.Conducted a user study measuring the usability and propensity for insight of fairness AI technologies in the workplace & Utilized findings to conduct a complete fairness assessment on a new company-wide machine learning effort.

With this experience, I am using analytical techniques to normalize and interpret proteomic data from diseased mice with different treatments. I will then use the techniques with the best results on multi-scale data to identify networks or biological processes influential in diseases and treatments. Through this process I have utilized a plethora of programs, including Treeview, Matlab, several packages in RStudio, and several statistical algorithms featured as add-ins on major applications.

I created a statistical pipeline that profiles Inflammatory Bowel Diseases, utilizing microbial samples from a plethora of locations within the gastrointestinal tract. I utilized a plethora of machine learning techniques to differentiate between diseases and identify outlying microbiota for successful pre-symptomatic disease prediction.

I began my first research experience at Princeton University with Dr. Anastasia Baryshnikova. We worked on creating data visualization software for mapping the yeast genome utilizing data collected on the yeast ‘phenome’. My task is data retrieval; For relevant datasets, I isolate variables important to the phenome and publish my results on a user interface, accessible by the entire scientific community. I have continued with this project with the hope that I can decisively correlate each gene in the yeast cell with a significant phenotype, potentially allowing me to determine a function/family for each gene based on the reactions seen in hundreds to genome-wide experiments.

I worked in a Data Science lab at the College of Charleston, with Dr. Paul Anderson and Dr. Andrew Shedlock. During this experience, my project focused on trying to computationally isolate transcriptional changes that distinguish relapsing lung-cancer patients from patients who do not relapse. I used multivariable, statistical methods to analyze the RNA-seq output from 21 lung cancer patients, 11 of whom were disease-free after treatment and 10 of whom relapsed after treatment. I compared the results from this approach to the univariate, traditional methodologies. I was able to isolate a list of isoforms that are differentially expressed between the two phenotypes, and further wet-lab tests will confirm the accuracy of my results.