Building a LLM from Scratch
INPROGRESS
About me
My technical skills include Python, C++, C, SQL, and various tools like Unix/Linux, Jenkins, GitHub, Snowflake, and AWS. I've interned at Toyota Financial Services, where I significantly improved data pipeline efficiency.
I've contributed to academic research in quantum network communication and developed a novel transfer learning method called BUS for efficient large language model training. With experience in data engineering, machine learning, and web development, I'm well-equipped to tackle complex challenges in software engineering and data science. My ability to communicate complex ideas effectively makes me valuable in both academic and industry settings.
Feel free to read my articles, experience, and projects below.
Articles
Work Experience
Graduate Teaching Assistant
Assisted in teaching a graduate-level Natural Language Processing course, covering foundational models to modern machine learning and AI architectures, wh...
Data Engineering Internship at Toyota
Achieved a 99.9% increase in efficiency by reducing time-to-delivery for new data pipelines from thousands of hours to minutes. Developed a Python scrip...
Teaching Assistant
Served as a Teaching Assistant for CNT 4104 Computer Networks Lab for IT, focusing on the OSI model and network protocol implementation. Guided students...
Research Assistant
Conducted innovative research on healthcare network infrastructure vulnerabilities, leveraging machine learning techniques to analyze and predict network ...
Web Developer
Designed and developed a responsive, multi-device compatible website for the USF Quantum Initiative using HTML, CSS, and JavaScript. Implemented ADA com...
Projects
EverywhereGPT
This project utilizes OpenAI’s API along with certain python libraries to allow you to use ChatGPT wherever you are typing. You can prompt an LLM at the push...
Bootstrapped Model UpScaling (BUS)
Developed BUS (Bootstrapped Model UpScaling), an innovative method combining nGPT architecture with fractional HyperScaling to enhance Large Language Mode...
Philosophical Similarity
Collected, cleaned, and analyzed a dataset of classic philosophical texts using GPT-2, demonstrating proficiency in natural language processing and data p...
NLP Model Improvement
Developed an academic machine learning replication paper that improved upon and scrutinized a state-of-the-art model’s performance on the Stanford Natural...
Deep Reinforcement Learning Agent
Led a team in developing a deep reinforcement learning agent for SuperTux Ice Hockey, implementing a Deep Q Network (DQN) algorithm using PyTorch. Achie...
Learning to Skim Text Transformer Replication
Fine-tuned an NLP model that surpassed previous state-of-the-art performance in text skimming tasks. Achieved significant improvements over the proposed...
Warehouse Inventory Management
Led a team in developing a comprehensive inventory management website for store owners, showcasing project management and leadership skills. Architected...
Image Classification with CNNs
Developed a Convolutional Neural Network (CNN) to classify images from the CIFAR-10 dataset, demonstrating proficiency in deep learning and computer visio...
Traveling Salesman Problem Solutions
Developed a comprehensive suite of 6 algorithms to tackle the Traveling Salesman Problem (TSP), a classic NP-hard optimization challenge: