My Projects

An end-to-end Python → SQL → Power BI pipeline answering a real civic question: is Kansas property tax growth driven by rising valuations or rising mill levy rates — and would relocating to a neighboring state actually lower the bill? Extracted and validated data from unstructured government PDFs (KDOR, Lincoln Institute) with pdfplumber, loaded it into a SQLite database with SQL views, then built an interactive Power BI dashboard with a county-level deep dive and a 5-state comparison.

25 years of NTSB aviation incident data (39,000+ records) explored through an interactive Streamlit dashboard — trend lines, injury-level breakdowns, weather-condition analysis, and a US choropleth heatmap of incident density. Built with the domain expertise of an FAA-certified Airframe Mechanic: every insight is grounded in real aerospace maintenance and regulatory experience, not just the raw numbers.

An end-to-end data analytics pipeline exploring 1,500 personal finance transactions across 5 years (2020–2024): synthetic data generation, data cleaning, SQL-based analysis in SQLite, interactive Plotly visualizations, and a 6-month linear regression forecast. Full workflow documented across 4 Jupyter notebooks covering cleaning, SQL queries, visualization, and forecasting, with dynamic KPI filtering and income/expense breakdowns by category and day of week.

A full-stack booking platform built for a real paying client — an esthetician who needed scheduling without a subscription fee. Custom appointment system with an approval workflow, service menu management, and a mobile-optimized admin dashboard. Requirements through deployment, owned end to end.

A Django REST Framework API implementing the full WHS golf handicap formula from user-submitted rounds — a free alternative to a $60/year official subscription. JWT-authenticated, with a separate React/TypeScript frontend and Stripe integrated for a planned premium tier with extended trend analysis by course and weather condition.

A real-time shared household finance tracker built around instant accountability — no bank linking, just an honest running balance based on what’s actually logged. Expense and balance updates push instantly to both partners via Django Channels and Redis, with graceful degradation built in: if a WebSocket connection can’t reach Redis, the app still works fully on standard page loads.

A full-stack golf community platform where golfers share round recaps, course conditions, and scores. Users log rounds, report live course conditions to help others plan visits, and browse a community feed — with object-level permissions ensuring users can only edit or delete their own posts.

A Django app that scores 10-day weather forecasts for deer hunters, ranking each day 0–105 based on five weighted factors (temperature, barometric pressure, wind speed/direction, precipitation, and moon timing) known to drive deer movement. Pulls live forecast data from Open-Meteo and calculates moonrise/moonset astronomically with ephem — no paid APIs required.

A full-stack budgeting app for tracking income and expenses, categorizing transactions, and visualizing spending by category. Includes summary cards, date-range filtering, and a Recharts breakdown of expenses by category — with a public read-only demo account so anyone can explore it without registering.

A retrieval-augmented generation pipeline: resumes are parsed with pdfplumber, chunked, and embedded into a ChromaDB vector store. The most relevant chunks are retrieved by similarity search against a job description to ground GPT-4o-mini’s skill-match scoring and generated cover letter in the candidate’s real experience — not general knowledge.

A bowhunting ballistics and arrow-building platform: real-world arrow speed estimates from bow specs, kinetic energy and momentum calculators, ethical-harvest threshold ratings by game animal, and a community-submitted arrow component database that powers a full arrow builder with automatic weight/speed/KE calculations.

A production-style AI agent built with LangGraph and Django REST Framework. Plain-English requests hit a REST endpoint, and the agent automatically selects the right tool (live weather, calculation, or an internal knowledge base), executes it, and responds in natural language — with full conversation memory across turns via session ID.