Here you'll find some of my personal development projects as well as educational and research work.
| Title | Completion Date | Description | Tools | Link(s) |
|---|---|---|---|---|
| Research - Temporal Reasoning in LLMs | May 2024 | Researched limitations of temporal reasoning with Large Language Models (LLMs). Fine-tuned different LLMs (BERT, Google FLAN-T5, Falcon 7B) through fine-tuning (masked, Q/A, graph generation) and compared results to original benchmarks. | Python (PyTorch, transformers, sklearn, pandas, NumPy, NLTK), Huggingface API | Reports |
| Pun Detection and Location | Dec 2023 | Created a program that detects, identifies, and explains pun words trough natural language processing (word similarity calculations, phonetics, transformers), LLM fine-tuning (masking, prompt engineering), and machine learning (SVM, LSTM) methods. | Python (sklearn, NLTK, pandas, NumPy), Llama 2 | Reports and Presentations |
| Revenue & Pricing Analysis | May 2022 | Modeled predicted revenue streams to highlight where reinvestments can be made, discovering inconsistencies in pricing of services across the consulting company. Pivoted to modeling service pricing to improve consistency of data for future modeling purposes. Leveraged time series forecasting as well as maching learning (random forest, XGBoost, resampling) methods. | R (tidyverse, dplyr/dbplyr, ggplot), RStudio | Final Presentation |
| Dropshipping Detection | August 2021 | Created a program that scans Amazon product links and searches for inputted items on bulk supplier websites (Alibaba). Experimented with natural language processing and web scraping technologies. | Python (NLTK, BeautifulSoup, pandas) | --- |
| Data Mine - Job Scraping | May 2021 | Scraped employment data from large online databases and classified people based on job type and seniority, with the end goal being targeting professionals with personalized job recommendations. Leveraged web scraping and encoding-based text classification methods. | Python (spaCy, pandas), JavaScript, Node.js, MongoDB | Final Presentation |
| Hello World Hackathon | October 2020 | Leveraged Google's Fact Checker API to create a fact checking application. Created a user-friendly web-app within the 24 hour time limit. | HTML, CSS, JavaScript | --- |