Software engineering portfolio

Web software that shows its working.

Browser applications and small Python services where correctness is the feature: exact money math, citations tied to the passage they quote, drawings that export identically every time, workflows whose every step can be replayed. The code is implemented with AI assistance (Claude Code) under my direction and review; each project comes with its architecture notes, automated tests and an acceptance review by a separate AI session.

Projects verified: 10
Published with live demos: 10
Separate AI review sessions: 10 of 10

Featured case studies

01 · Published · 2026-10-15
ProfitLab
Marketplace unit economics with exact decimal arithmetic and pass/fail gates.
A reseller deciding what to list needs projected profit per product after every fee, shipping cost and overhead share, and a clear yes/no against minimum profit and margin. Rounding drift or a blank cost silently treated as $0 makes a losing product look like a winner.
Hard part: Proving the money math.
TypeScript · React · Vite · decimal.js · zod · Vitest
- Read the case study: ProfitLab
- Live demo of ProfitLab
02 · Published · 2026-10-15
RecruitRadar
Hiring-market intelligence from public job postings, with an explainable score.
Public job postings are noisy: the same company appears under several names, roles are reposted, and a single number hides why a company looks active. A recruiter needs normalized postings, careful de-duplication and a score whose every part can be checked.
Hard part: Two implementations that must agree.
TypeScript · React · Vite · Python · FastAPI · SQLAlchemy
- Read the case study: RecruitRadar
- Live demo of RecruitRadar
03 · Published · 2026-10-15
SourceTrace
Claim and citation review with exact passages and a human-only approval gate.
Research notes drift away from their sources: a quote gets paraphrased, a citation points at the wrong paragraph, and nobody can say who approved a claim. Reviewers need every claim tied to the exact passage that bears on it, and a record of every decision.
Hard part: An audit trail that cannot break the workspace.
TypeScript · React · Vite · zod · Python · FastAPI
- Read the case study: SourceTrace
- Live demo of SourceTrace
04 · Published · 2026-10-19
SceneForge
A deterministic procedural SVG studio for hand-drawn looking diagrams.
Explainer videos need simple diagrams (stick figures, arrows, regions) that look hand-drawn, can be edited, and come out identical every time they are exported, so a revision does not change the drawing everywhere else.
Hard part: Identical output, every time.
TypeScript · React · Vite · zod · SVG · Vitest
- Read the case study: SceneForge
- Live demo of SceneForge

More projects

Each passed the same build and separate review process and is published with a live demo.

Workflow Observatory
A state-machine studio: approvals, retries, idempotency keys, replay and comparison.
Published · 2026-10-19
SignalScope
Educational historical event studies with no look-ahead.
Published · 2026-10-19
ScriptAudit
Four deterministic audits for video scripts, with suggestions accepted one at a time.
Published · 2026-10-19
TopicPulse
Explainable topic scoring from public counts, relative to each channel.
Published · 2026-10-19
RecruitSim
Branching recruiting practice with a transparent rubric.
Published · 2026-10-19
StoryAtlas
An explorer for a fictional world, with a spoiler filter and content warnings.
Published · 2026-10-19

How the work is done

Testing

Each project ships unit tests and browser tests of its main user journeys against the production build, and most start from a written specification. Where the logic matters most, a second implementation checks the first: a BigInt reference model for money, two independent engines held together by golden vectors and fuzzing, byte-for-byte golden SVG output, exact replay of an event ledger.

Independent acceptance

After each build, a separate Claude Code session that did not write the code re-runs everything from a clean clone, tries to break the main journeys, and fixes what it finds with regression tests. It is an AI review, not a human code audit; the findings and the commands it ran are listed in each repository’s test report.

Data validation

Imported files, stored workspaces and API responses are treated as untrusted: strict schemas, size limits, line-numbered errors, and missing values kept missing instead of becoming zero.

Responsible AI

The specifications, code, tests and documentation were written with Claude Code, Anthropic’s coding agent, under my direction and review; none of it is presented as written by hand, and each repository says so in AI_ASSISTANCE.md. Inside the apps, AI is optional and off by default; where it exists it only suggests drafts that a person must accept, and nothing claims to verify truth or predict outcomes.

Stack, as used in these projects

Skills and where the code shows them
Area	Tools and techniques	Where to see it
Frontend	TypeScript (strict), React, Vite, accessible HTML and CSS, SVG	All ten projects; SceneForge and SignalScope draw their own SVG.
Backend	Python, FastAPI, SQLite, SQLAlchemy and Alembic, Pydantic	Optional local services in RecruitRadar (SQLAlchemy and Alembic), SourceTrace, ScriptAudit and TopicPulse (SQLite with ordered migrations).
Testing	Vitest, pytest, Playwright, golden vectors, differential fuzzing, reference models	Every project has a one-command verification script that fails loudly.
Data handling	zod and Pydantic schemas, CSV import with line-level errors, exact decimal money, versioned local storage with recovery	ProfitLab, RecruitRadar, SourceTrace, SignalScope, TopicPulse.
Delivery	GitHub Actions CI, GitHub Pages, static builds under a sub-path	All ten projects are published with CI and GitHub Pages.

Contact

The best way to reach me is through GitHub: github.com/PatrickNDevelopment.