Tuan (Kevin) Le (He/Him) is an undergraduate at DePauw University specializing in Computer Science and Mathematics. He joined Vanderbilt's Open Ending Learning Environment lab as an undergraduate research assistant through The Leadership Alliance and subsequently became a research assistant in the Data Science Institute at the same institution. During the 2023 summer, he assisted research in dissecting the Large Language Model's architecture to explore the performance of a model with different structures, specifically for grading scientific short answers. He also helped develop an evaluation scheme for the models' capacity in grading, which is used to assess the practicality of the subsequent project. He is fascinated by the concept of artificial intelligence, especially the connection between the artificial and biological brain in understanding and perceiving conceivable aspects. From vision to language, the world is such an oyster for AI to explore. It is interesting to develop AI in ways that can eventually constitute thoughts to carry out human-supervised tasks, tasks that are impractical to be handled by humans. I have a vision that in the near future, we will be able to allow robots to perform work too cumbersome or difficult for humans physically, like exploring the ocean, and space, or helping victims of construction failure or fire. He plans to pursue a doctorate education either in Computational Neuroscience studying the bridge between the artificial and biological brain to develop a foundational understanding of the understanding and creation of language or in computer vision, specifically developing human perception for robotic devices, in the hope of modifying connection between senses like touch, sight, or smell with the decision-making process of robots. His research involves mathematical analysis, computer science, and physics, and works on classifying moving patterns of longboards into their main categories using 3D acceleration vectors. With each pattern having collapsing acceleration patterns, the challenge is to extract the right features, from frequencies to paths, write a sliding window approach for it and work with high-dimensional data for multi-layer neural networks. He utilizes compressed frequency information collected in time windows to identify actions in those time windows since longboards’ actions are not distinguishable per second. I utilized multiple classifiers from Gradient Boosting to K-Nearest Neighbor, while also developing recurrent and 1D convolutional deep neural network models for classification. Such work was produced to help with the training of the sport. This work has laid the foundation for my basic understanding of artificial neural networks and grown my interest in pursuing further studies in this largely theoretical yet applicable field. His research also involves developing prompts for different domains, from creating text-based games to language-learning materials. He has also done research on recommendation systems leveraging transformers for social media and online service providers.