Data Science Portfolio
Turning messy data into decisions businesses can act on.
About
I'm a data science graduate who likes turning messy data into insights businesses can act on. My coursework covered Python data wrangling and visualization, NLP (custom NER, sentiment analysis, deep learning embeddings), statistics, and business intelligence design.
I also have HTML, graphic art and automation skills, so I'm comfortable owning a project from idea to completion. I'm looking for a data science role to bring business solutions to the table.
Projects
Each project below shows the problem, my part in it, and what the data said.
Can the characteristics NOAA records about a storm tell us whether it will cause economic damage? Two models tested on five years of Storm Events data. My role: data visualization.
Forecasting hourly grid demand from EIA-930 data. My role rotates to data wrangling for this sprint. Full case study goes live at the end of Week 8.
Predicting contamination risk in municipal recycling streams from EPA MSW data. My role rotates to data scientist for this sprint. Full case study goes live at the end of Week 11.
Skills
Coursework
Code, output and write-ups from my data science courses, rendered so you can read them in the browser.
Published R Markdown report on screen time addiction, with analysis and visualizations.
Cleaning and exploring the Digital Habits vs. Mental Health dataset, then a final project with charts that show what drives the outcomes.
Trained a custom named entity recognizer on a real PubMed Central open-access corpus, with evaluation.
Hotel-review sentiment two ways: VADER plus logistic regression (methods adapted from Sarkar), and a Keras deep learning model on Universal Sentence Encoder embeddings.
Classifies Amazon Alexa reviews as positive or negative using TF-IDF features, with full evaluation.
Bayes’ theorem on a manufacturing probability tree, plus a custom Shannon entropy function compared across distributions.
More work
Case work from my BI coursework, plus a document-processing automation I built in Power Automate.
Baby Bio written case solutions with a full data architecture diagram covering ingestion, ETL, storage, ML and BI layers.
Business case and requirements for a digital wellness analytics platform: objectives, scope, timeline, budget, ROI, and stakeholder interviews.
Trained two AI Builder document models to extract invoice fields, tested them on sample invoices, and ran an automated flow end to end. Left: choosing the fields to extract. Right: the successful run.
Contact
I'm looking for a data science role where I can bring business solutions to the table. Send me a note.
HaliSarahParsons@gmail.com