§02ProjectsAs-built
Selected work
Open any project for the write-up and, where there is one, a demo you can run here in the page.
TaskManager REST APILiveLive demo
Containerised REST API with EF Core and PostgreSQL, key-protected writes, and a signed image that my server deploys automatically.
Open →02Resume Classifier APILive
Sorts resume text into five job families with TF-IDF and logistic regression, served as an API from my home server.
Open →03aoda-scan
A command-line tool that crawls a whole website and grades it against WCAG 2.1 AA and Ontario’s AODA.
Open →04AgentMesh
A self-hosted control plane for running AI coding agents across several model providers, where every user brings their own keys.
Open →05Home Server
An old laptop run like production: Cloudflare Tunnel, signed pull-based deploys, monitoring and tested backups.
Open →06Train Yard Management SystemLive demo
Rail inventory and safety validation in C. Enforces weight limits, locomotive pull capacity and car-type protocols, with a test suite driving the same logic layer.
Open →07ArenaCore RPG EngineLive demo
C++ engine built around an abstract combatant hierarchy, applying the Rule of Three, operator overloading and manual memory management.
Open →08Inventory CRUD
Category and supplier management on ASP.NET Core MVC — Razor views, view models, and EF Core migrations against SQL Server.
Open →09This Portfolio
The site you are reading. React and Vite, a hand-built CSS design system, deployed to GitHub Pages by an Actions workflow on every push.
Open →§03Also built
Smaller pieces
C Projects
Baby name popularity search over census CSVs, and a train inventory console app.
View repository ↗—C++ Exercises
Marketplace, credit card validation, restaurant ordering, sorting, and a lexical store engine.
View repository ↗—C# Fundamentals
Console applications covering OOP basics — bank simulator, library manager, grade tracker.
View repository ↗—Shell Scripts
Utility scripts for development workflow automation.
View repository ↗—AI Programming Tools
Notes and references on prompting, neural network fundamentals, and software licensing.
View repository ↗Resume Classifier API
Sorts resume text into five job families with TF-IDF and logistic regression, served as an API from my home server.
A scikit-learn pipeline (cleaning, TF-IDF features, logistic regression) served with FastAPI. It trains on a generated corpus, because real resumes are personal data.
The first generator gave each role its own words and the model scored a perfect 1.00, which measured the dataset rather than the model. The corpus now shares filler and tools across fields, and 45% of resumes borrow a line from another field. Accuracy on generated text is about 0.98: proof the pipeline works end to end, and nothing more.
Every label comes with a confidence. Unrelated text scores about 22%, barely above the 20% chance baseline, which is the model saying it has no idea.