Defeating ALPR
Data-science capstone studying adversarial robustness of Automatic License Plate Recognition systems using evolutionary optimization.
- Problem: ALPR systems track vehicles and share data with law enforcement regardless of suspicion.
- Approach: Built a perturbation pipeline and evolved minimally visible attacks using NEAT.
- Findings: Achieved 100% evasion; effectiveness is plate-specific with no universal pattern.
Computer Vision
Machine Learning
NEAT
Python
Privacy
Gym Analytics Dashboard
A live, data-driven dashboard tracking my workout progression — pulling from a GitHub Actions pipeline that automatically syncs my gym data.
- Problem: Gut-feeling training leads to stalled progress. I wanted the data to tell me what's working.
- Approach: GitHub Actions exports my Google Sheets log to CSV; the dashboard renders SVG charts with zero dependencies.
- Outcome: Surfaces volume trends, exercise PRs, streaks, and muscle-group imbalances every session.
ETL
GitHub Actions
Dashboards
SVG
Supplier Analysis
Analyzed revenue by supplier, lead times, and the tie between delivery speed and financial performance.
- Problem: Identify top suppliers and bottlenecks across the chain.
- Approach: EDA with SQL/Python, feature pivots, and visualizations.
- Insight: Clear pattern between on-time delivery and higher revenue share.
PythonSQLSupply Chain
TFT Level Team Search Tool
A front-end tool to search TFT team compositions by level and already-owned units, ranking results to minimize decision time.
- Problem: Quickly identify the best comp to pivot into based on current board state.
- Approach: Level-based filtering with a "what do you already have?" selector; comps scored by unit overlap.
- Outcome: Surfaces the most achievable comps first while showing missing units and active bronze traits.
PythonHTMLCSSJavaScript
XGBoost Home Regressor
Used Python to extract, clean, and apply machine learning to predict housing prices.
- Problem: Estimate home prices based on complex market features.
- Approach: Data cleaning, feature engineering, and XGBoost regression.
- Outcome: Strong predictive accuracy and feature insights for key housing factors.
PythonXGBoostRegression
Job Salary Predictions
Cleaned job-market data and built regression models to predict salaries.
- Problem: People are often underpaid — how do I know what I should make?
- Approach: Cleaned a large dataset and applied linear regression.
- Insight: Strong signal from experience level and company size; interpretable trends.
PythonPandasRegression