I build the machinery that makes software teams faster and safer: developer experience, CI/CD and quality gates, platform tooling, and AI that actually ships to production.
$ whoami
christoffer maintz - developer, aarhus dk
$ cat focus.txt
dx · devops · backend · platform · ai enablement
$ ls ~/shipped
illux-product-ai foundry jev-tools leash jobbuddy tech-atlas
$ gh repo list --public | wc -l
$
$ git log --all-repos --latest
synced 8 Oct 2026 · time since last push
$
Fast feedback loops, sane defaults, tooling that removes friction instead of adding ceremony.
Read more →Reusable pipelines, deterministic quality gates, supply-chain and secret scanning.
Read more →APIs and data models built around the domain, with queues and retries where dependencies are slow.
Read more →Paved roads: pinned toolchains, scaffolding, config-as-code and policy-as-code.
Read more →LLM features in production, agentic dev workflows, and guardrails that fail safe.
Read more →Generative-AI product enrichment & room visualization for Shopware 6
A production Shopware 6 plugin that turns product images into multilingual copy, SEO, categories and tags with generative AI, and lets shoppers see artwork on their own wall before buying. Designed, built and deployed as sole developer.
A polyglot engineering gate for AI-assisted development
One deterministic quality standard - format, lint, types, tests, coverage, security - enforced identically while an AI agent edits, before push and in CI. Reusable GitHub Actions workflows, mise templates and a Claude Code plugin, released as versioned open source.
Three TypeScript tools that put a cheap, typed model in front of the expensive one
A monorepo of tools built on TypeSafe AI's Jev - an agent tool-call guardrail, a GitHub issue-triage Action and a bulk classification CLI - that run Jev on everything and hand only the uncertain cases to a human or a real LLM.
Keeps a coding agent to the project rules no linter can check
A Jev-powered guardrail that judges every agent turn against the un-lintable rules in your CLAUDE.md or AGENTS.md, and tells the agent exactly which one it broke so it fixes it before moving on. About 300 ms and a fraction of a cent per turn.
AI job-application platform - from posting to tailored CV
A Spring Boot (strict hexagonal) + Angular platform that ingests job postings and generates tailored CVs and cover letters with ATS match reports, backed by Postgres full-text and pgvector semantic search.
A bilingual technical dictionary you can explore as a map
About 420 security, CS, AI and platform terms in English and Danish, linked by typed, sourced relationships into a knowledge graph with a 2D/3D explorer, plain-language explanations and quizzes.
One React codebase, running on the web and as an LG webOS TV app
A movie and TV browser that pulls titles from TMDB, adds IMDb and Rotten Tomatoes scores, shows where each title streams in your country and, on the TV, launches straight into the right streaming app. Same React tree for mouse and remote.
Spin a wheel, settle movie night
Filter, spin a 12-slot wheel and get a title plus where it streams. React on Cloudflare Workers, with the TMDB key kept server-side behind a rate-limited proxy that checks origin and path.
TypeScript, Java, C#, PHP, Kotlin, Swift, Python, Spring Boot, Symfony, Angular, React, Astro, GitHub Actions, mise, Docker, PostgreSQL, RabbitMQ, Cloudflare, Azure, Kubernetes, Claude Code, Semgrep
Why I stopped treating CI and "AI coding standards" as two separate things - and what Foundry does instead.
A new home for my work, my ramblings and rants, and the longer versions of things I post elsewhere. Plus a look at what I built under the hood.