Large language models (LLMs) such as ChatGPT, Claude, Gemini, and Llama are increasingly being deployed as search and decision-support tools for health-related inquiries. As users provide demographic context, including age, to obtain personalized guidance, these systems can differentially adjust tone, directive strength, or safety framework. Although age can be clinically relevant, unintended variation in the generated advice can introduce systematic safety disparities or representational bias. In this study, we analyze outputs from two major LLM families across 10,679 physical and mental health scenarios to examine how explicit age cues, including teen, young adult, middle-aged, and senior, influence the safety and linguistic properties of generated health advice. To quantify these effects, we introduce three task-specific evaluation metrics: Age Differential Safety Bias (ADSB) to measure relative safety shifts under demographic conditioning, Safety Risk Score (SRS) to capture cumulative weighted unsafe recommendations, and Tone Differential Index (TDI) to detect systematic changes in linguistic complexity and formality associated with representational harm. The results indicate that explicit age cues systematically alter the behavior of the model. Demographic conditioning consistently reduces safety quality relative to age-neutral baselines. Middle-aged cohorts exhibit a higher cumulative safety risk in directive responses, whereas senior cohorts demonstrate elevated tone shifts consistent with oversimplification and increased formality. These findings suggest that demographic sensitivity can introduce measurable allocative and representational disparities in healthcare-oriented LLM systems. This work establishes a reproducible audit framework for evaluating demographic safety sensitivity in health-focused LLM deployments and contributes to the development of standardized evaluation protocols for safer and more equitable integration of AI systems in clinical and consumer health environments.

The BatLab project aims to develop a machine learning based tool that assists researchers in identifying bat species from acoustic recordings. Bats rely on echolocation calls that vary in frequency, duration, and shape, allowing species to be distinguished through analysis of their recorded calls. Currently, researchers must manually review large volumes of acoustic recordings, which is a time consuming process that limits the scale of ecological studies. This project explores the use of supervised machine learning to automate the classification of bat echolocation calls using labeled training data. The system analyzes acoustic features within recorded calls and predicts the most likely species while flagging uncertain cases for further review. In addition, the project focuses on improving data organization and providing a user friendly interface that allows researchers to efficiently visualize and manage acoustic data. By reducing the manual workload involved in analyzing bat call recordings, the BatLab system aims to support ecological research and improve the efficiency of studying bat populations.

This senior design project focuses on developing PostAgent, an AI-powered content creation platform created by Corevation, an innovations tech company. This product is aimed at helping businesses and entrepreneurs with creating and managing social media content more efficiently and allow marketing endeavors to be more manageable. Our team is building multiple features, including AI tools to regenerate and edit post text and images, an analytics dashboard for tracking social media performance, and a centralized content library for organization purposes and for users to upload custom content to the platform. We are also assisting in the overall UI/UX to ensure an intuitive user experience and developing a company website to support Corevation’s public presence. Together, these components demonstrate a full-stack approach to product development, blending AI capabilities with user-centered design.

How would one classify an apple fruit versus an apple phone? Typically as a fruit and a technology device. However some modern systems for classifying patents are insufficent and would be unable to differentiate between the two and cluster both based on their containing the word ‘apple’. Our task with iPELiNT is to build upon solutions to better visualize how USPTO( United States Patent and Trademark Office) art unit’s change over time. An art unit is a group of USPTO examiners specializing in a specific technology area. Our end product helped establish a data-driven system for conducting forensic analysis of USPTO patent examiner dockets using vector embeddings and internal data pipelines. We used mongoDB for our database, JavaScript and Python for our backend, and NuxtJS and Vue for our frontend. Our 5 phases of development are as follows. 1. Data Aggregation and Preparation. 2. Centroid Calculation and Art Unit Profiling. 3. Deviation Analysis and Scoring 4. Visualization and interpretation Framework.

Artificial intelligence tools, especially large language models (LLMs) are progressively being integrated into educational settings as resources that can enhance student learning and offer novel methods for information retrieval. As these technologies advance, educators and researchers are increasingly focused in comprehending their impact on student learning and engagement with academic content. This study examines the potential role of AI-based systems in facilitating student learning by analyzing various ways employed by students to obtain and process information during study activities.
The study's participants are split up into four groups, each of which accesses learning resources in a different way. The first group relies on traditional text-based study resources. The second group uses general online resources to gather information. The third group is allowed to use AI-based tools powered by large language models to receive explanations and assistance. The fourth group uses a hybrid strategy that blends AI-supported tools with conventional study materials.
The performance and learning experiences of these groups are compared to evaluate how different resources influence students’ understanding of course concepts. The findings are expected to provide insight on whether AI technologies can successfully supplement conventional teaching methods. Understanding these effects help educators determine how to appropriately incorporate AI and LLM tools into classroom settings to improve learning while upholding efficient teaching methods.