For this position, I conducted a literature review for topics such as large language models, causality, and disentangled deep learning. I then was able to apply these concepts and others such as transformer self-attention to create custom models for the purpose of analyzing chemical compounds and their effects on cells. I'm also collaborating to design structures capable of thorough creation and evaluation of models in hyperparameter sweeps in order to collect statistics for optimal model performance.

During the Fall 2022-23 semester, I was a teaching assistant for CMPINF 401 - Intermediate Programming, in which I taught students about introductory concepts to computer science in the Java programming language. I proctored weekly labs with programming assignments which I verified and graded. Such topics covered in the course and in Labs include data abstraction, encapsulation, arrays, exception handling, and recursion. During the Spring 2022-23 term, I was a teaching assistant for CS 449 - Introduction to Systems Software course, in which I led weekly lab sections where I taught students about lower level concepts in the C programming language. This course was meant to teach students about memory management details, compilation, multiprocessing, and assembly, as well as other concepts.For the current Fall 2023-24 semester, I am a teaching assistant for CS 1501 - Algorithms and Data Structures II, where I will be assisting students with concepts and projects related to data structures such as trees, heaps, and graphs, as well as compression algorithms, search algorithms, and clustering.

For this position, I analyzed the applications of the transformer and BERT-like models for the task of protein structure prediction. The models were constructed using PyTorch and PyTorch Lightning, and received an amino acid sequence in an attempt to calculate its folded, three-dimensional structure. I had to read and analyze machine learning literature in an attempt to recreate specific models to use in this task, such as the universal transformer model.

For this position, I was responsible for reviewing and replication machine learning and neuroscience research to analyze proper methods for preparing electroencephalogram (EEG) deep learning classification models. This included analyses using spectral images and convolutional neural networks (CNNs), as well as repurposed transformer models for analyzing EEG signal sequences. This led to a collaborative effort on a paper discussing these methods and practices pending journal submission.

In this role I developed a library designed for converting raw electroencephalogram recordings into spectral images for the purpose of constructing convolutional neural network pipelines for classification tasks, such as determining probabilities of patients having neurodegenerative diseases. I was also given a deep learning classification task in which I used the mentioned library and other creative methods in an attempt to perform an analysis using a small subject dataset. This turned into a report using descriptive statistics to determine how the model could perform in a clinical environment, as well as discussions as to what could improve results and make the model architecture more feasible in a real world setting.

In this position, I was responsible for grading weekly math assignments for all students at the learning center, from Kindergarten through High School, grading and teaching topics ranging from addition and algebra to advanced topics in calculus.