Work

Selected proof from the kind of systems I like building.

I use project examples to show the shape of the work: what was built, why it matters, and how it helps people do better work.

Project

AIWAR

The rebranded successor to ahref: an SEO and site-audit tool rebuilt around a clearer identity.

Shows the ability to turn a working product into something more coherent without losing the technical core.

Project

AI-Dev-Team / Mission Control

A structured AI operating setup for planning, tasks, and delegated work across projects.

Useful proof that I can design systems around people, process, and repeatable execution.

Project

OpenClaw customizations

Workflow and tooling changes that make the assistant setup more durable and less brittle.

A good example of practical AI tooling work that stays focused on shipping instead of hype.

Project

Privacy-first measurement

Analytics and experimentation work that respects consent, data quality, and real-world constraints.

Still important — just no longer the headline identity.

Deep-dive areas

The older site had more detail. This keeps the useful parts.

The aim is to keep the practical expertise visible without dragging the site back into the old analytics-first identity.

Matomo analytics & tagging Clean event models, migrations from UA/GA4, consent-aware tracking, and data quality are still core parts of my toolkit.
SEO, AEO & GEO Technical SEO, structured data, knowledge graphs, and answer-engine optimization are all part of the same discoverability problem.
Experimentation & optimization Hypothesis-driven A/B tests with guardrails, proper statistical planning, and a focus on what changes behavior.
Data ops & automation Pipelines, dashboards, automated reporting, JS, Python, and Docker all help when the work needs to keep running.
How I think about work

Useful work has structure, not just features.

The best projects usually combine clarity, enough restraint, and a path to maintenance.

Define the real problemMake the system serve the use case, not the other way around.
Keep the signals honestMeasurement should survive real traffic, consent, and edge cases.
Ship with a maintenance storyIf it can’t live with one person, it’s probably too big for the job.