During my Software Engineering Internship at Syntax, I worked on the Generative AI team, where I deployed Retrieval Augmented Generation (RAG) systems for large enterprise clients, optimizing data retrieval with FAISS and ChromaDB vector stores. I engineered knowledge bases using client-specific data and OpenAI Embeddings, enhancing the contextual relevance of AI-generated outputs. Additionally, I designed and integrated a custom LangChain tool that allows users to download AI-generated content in various formats, greatly expanding the system’s functionality. To ensure user safety, I also implemented real-time streaming and LLM Guard for query sanitization, contributing to a secure and engaging user experience.

As Recruitment Chair for Tau Epsilon Kappa, I led a team of 14 committee members to orchestrate a successful recruitment season. I managed logistics for four large open events and two closed events, and through strategic planning and targeted outreach, I increased event attendance to over 200 prospective members. My efforts culminated in the induction of 20 new members after a rigorous three-round selection process, including individual interviews for finalists. This role honed my leadership, organizational, and communication skills as I guided the team through a comprehensive and selective recruitment process.

At Walmart, I worked on the Digital Experience Scoring (DEx) team, which works to provide critical insights about product health to product owners throughout Walmart. While on the DEx team, I developed a scheduled Node.js program to automate data collection for 20+ product teams. This automation, integrated with multiple APIs and BigQuery tables, significantly improved data accuracy and efficiency, reducing hours of manual work each month. I refined my strategy for my automation initiative after presenting to the Architecture Review Board, which is comprised of over 80 engineers to ensure adherence to high standards of work. I also played a vital role in onboarding new product owners, ensuring seamless adoption of DEx across the division. My work culminated in presenting strategic insights to senior stakeholders, including the VP of the Associate Digital Experience division, setting the stage for future automation initiatives for the DEx team.

While on the Michigan Data Science Team, I developed a classification model using convolutional neural networks (CNNs) to classify images as either dogs or cats. I expanded this project to include a dog breed classifier, ultimately achieving over 60% accuracy in dog breed prediction with transfer learning. I also built a Discord Chatbot employing advanced Natural Language Processing, including LSTM networks and transformers, and deployed it using Amazon Web Services (AWS). I worked with the SQuAD2.0 (Stanford Question Answering) dataset as training data, and I focused on utilizing this data to train the chatbot on a wide variety of topics.

In my research, I studied the intricacies of machine learning interpretability, a crucial yet challenging aspect of AI. My work primarily focused on examining the limitations of standard interpretability frameworks in the industry. I generated detailed interpretability visuals for a machine learning model, employing tools such as SHapley Additive exPlanations (SHAP) and Generalized Additive Models (GAMs). An integral part of my research involved writing Jupyter notebooks to assess participants' use of interpretability tools. These notebooks were distributed on a large scale through a Qualtrics survey, facilitating comprehensive data collection. At the conclusion of this project, I co-authored a paper, which will be published in PACM HCI (Proceedings of the ACM on Human-Computer Interaction).

In my role as a Film Production Assistant, I worked to ensure the seamless operation of commercial video productions throughout New England. My responsibilities included the transportation of film equipment between various locations, ensuring everything was available where and when it was needed. I closely communicated with directors and producers, coordinating the proper placement of lighting and cameras on set. This role not only required physical management of equipment but also involved ensuring client satisfaction and comfort during shoots.