Data Scientist Portfolio

Resume: My Data Scientist Resume

Proficiency: Python, R, Tableau

Familiarities: SQL, Jes, Power BI

Tech Stack: pandas, NumPy, SciPy, seaborn, matplotlib, statsmodels, tidyverse, Git, Jupyter, Quarto, Excel

---

Education

• M.D.S., Data Science – University of Pittsburgh (Aug 2025 – Present)
• B.G.S., Mathematics – Ball State University (Aug 2020 – May 2024)
• A.A., Computer Science – Ball State University (Aug 2020 – May 2022)

---

Work Experience

Lead Analyst – Iconic Care Inc (June 2025 – Present)

• Created financial forecasts identifying $75,000+ in projected savings using data modeling and visualization.
• Deployed predictive models with 90% accuracy and automated workflows to accelerate product readiness/improve decision-making.
• Automated ordering processes, reducing operational time by 50%.
• Designed dashboards that increased procurement efficiency by 37%.
• Uncovered $30,000+ in unpaid CGM patient balances using Python.
• Conducted trend analysis on 100+ diabetic supplies/solutions to identify new revenue channels.

Billing/Revenue Specialist – Iconic Care Inc (May 2025 – June 2025)

• Developed pricing tables and consignment structures that increased billing success by 65%.
• Integrated eVendor and drop-shipping into Brightree, cutting delivery costs by 30%.
• Built product “kits” to reduce cycle time by 25% and improve profit margins.

CEO / Data Scientist – Palmer Projects (May 2023 – Present)

• Collaborate with nonprofit and commercial teams to improve decisions through custom data tools.
• Build and deploy predictive models aligned with business strategy.
• Share findings through a public website blog to support the Data Science community.

---

Projects

MyCaddy | Physics-Based Golf Shot Calculator

Repository
Launch the Web App

About:
Multi-interface, physics-based shot distance calculator that adjusts carry based on wind, temperature, lie, and weather drag.
Python · Flask · Data Modeling · Web App & Desktop GUI

Overview

Simulates environmental effects on golf shots. Built for both beginners and professionals to make data-informed club selections.

Key Techniques

• Vector-based wind modeling
• Temperature-density scaling (±0.3% per °F from 70°F)
• Weather drag (e.g., rain: −4%)
• Lie severity carry distance penalties (up to 40%)
• Flyer lie simulation (88–95% confidence)

Tech Stack

• Backend: Python, Flask, Gunicorn
• Frontend: Jinja2, CSS, tkinter
• Hosting: Render

Impact

• Improved carry prediction by 20–30% in rough lies
• Reduced wind-related shot error by 7–12 yards
• Corrected temperature-based carry by up to 5 yards
• Increased club choice accuracy by up to 15%

---

Google Capstone: Salifort Motors | Employee Attrition Prediction

Repository
View Certification

About:
End-to-end machine learning project to help predict and prevent employee turnover using HR data.
Python · scikit-learn · pandas · XGBoost · EDA

Overview

Analyzed 15,000 records to build classifiers that predict whether an employee will leave. Generated HR-ready insights to support retention initiatives.

Key Techniques

• EDA for trends, correlations, and outliers
• Logistic Regression (baseline)
• Tree-based models (Random Forest, XGBoost)
• Performance evaluation (accuracy, recall, F1, ROC)
• Feature importance analysis

Tech Stack

• Data Handling: pandas, NumPy
• Modeling: scikit-learn, XGBoost
• Visualization: seaborn, matplotlib
• Environment: Jupyter Notebook

Impact

• Achieved 94% accuracy using XGBoost
• Top predictors included: satisfaction level, evaluation score, and number of projects
• Delivered actionable insights to reduce turnover risk
• Built a scalable retention modeling framework

---

CGM Billing Analytics | Python Excel Automation

Repository

Tools: Python, pandas, Excel, OpenPyXL

Overview

Script automates billing insights saving 200+ hours by filtering HCPCS codes and calculating patient responsibility from Excel exports.

Features

• Filters for codes A4239 and E2103
• Computes responsibility as (Allowed – Payments)
• Generates dynamic summaries and Excel sheets
• Consolidates financial trends for billing teams

Impact

• Identified $317,000 in unpaid patient responsibility
• Found 44% of CGM patients had unpaid balances, saving the company thousands
• Highlighted $245,000 still uncollected
• Diagnosed a 56:100 payment success ratio

---

Real Estate Price Modeling | Ames, TX

Repository

Tools: R, tidyverse, GLM, yardstick, ggplot2

Overview

Used 2,300+ observations and 60+ variables to build a predictive model of home sale prices in Ames, TX.

Key Techniques

• Data wrangling with tidyverse
• Log-transformation to normalize skew
• Predictive modeling via GLM
• Evaluation using yardstick::rmse_vec()

Impact

• Explained over 88% of the variance in housing prices using a Generalized Linear Model (GLM) applied to the Ames Housing dataset, demonstrating strong model fit and predictive reliability.
• Identified above-ground living area and neighborhood as the top two drivers of sale price, contributing to more than 60% of total model influence, enabling targeted pricing strategies for real estate professionals.
• Revealed that a finished basement adds ~$10K–$20K in value on average, providing actionable insights for investors prioritizing renovations.
• Empowers non-technical stakeholders with interpretable visualizations (coefficient plots, residual diagnostics), bridging statistical output and business decision-making.