Crystal Hollis

Automated Script Geocoding List of Locations

This Python automation script reads a CSV list of Hope Supply Co.’s partner organizations, which originally lacked full addresses. Instead of manually searching for each location, I automated the process using ArcGIS’s geocoding APIs to convert names into spatial coordinates. Developed during GIS Programming (GISC 2335), this project demonstrated how geospatial scripting can drastically streamline data preparation for mapping.

Lesson learned: how to automate geocoding workflows using Python and external APIs to efficiently transform incomplete real-world datasets into usable GIS layers.

SEO Keyword Generator

Proof of Concept on Hugging Face

I am developing a keywording tool with Python, Gradio, and Gemma LLM that streamlines photo/video metadata by combining controlled taxonomy enforcement with SEO signals (Google Autocomplete, Google Trends, Datamuse, WordNet) and outputs structured IPTC fields ready for metadata editing workflows. The proof of concept on Hugging Face is what I developed as part of my proposal while contracting for Southern Methodist University (SMU)'s Photo/Video Department.

Lesson learned: I learned how to wire Python functions directly to inputs/outputs, making a working prototype much faster than a custom Flask/Django build.

Maternal Mortality Analysis

This Jupyter notebook attempts to predict maternal mortality outcomes using machine learning models built with AutoGluon. While the project contains early mistakes and areas for improvement, it represents a meaningful effort to apply techniques learned in Python Data Science (ITSE 2371) to real-world health data explored in Personal/Community Health (PHED 1304).

Read the full research paper (PDF)

Lesson learned: how to bridge public health and data science concepts, and the importance of iteration, data quality, and thoughtful modeling in addressing critical health disparities.