The problem

Job hunting eats hours: postings are scattered across sites, and each one really wants a tailored CV.

What I did

Job Hunt Engine is a custom app, not a no-code workflow. It gathers the roles I care about from the main job sites in one place. I paste a description and Claude does two things in a row: it scores the fit with a traffic light, then pulls the right experience from my master CV and writes the best CV for that role. I only update the master and always have the right CV.

The result

I update a single CV and always have the right one, without redoing it every time.

The problem

Prototyping an interface takes development time, even just to try an idea.

What I did

With UIGen I describe the component in words and Claude writes the React code, with a live preview as it takes shape. I change a sentence and the code updates in streaming. All in a custom app, no files on disk and no setup.

The result

From idea to a working prototype in minutes, just by talking.

The problem

Between skiing, tennis and football, plus gym, diet and work, keeping it all under control and fitting it into daily life is a nightmare.

What I did

ATLAS goes beyond a dashboard. A knowledge base collects the work of the best trainers and nutritionists; Claude cross-references it with my medical history and builds tailored training and nutrition plans, finding the trade-offs between several sports and everyday life.

The result

From manual juggling to a coach that proposes plans calibrated to me.

The problem

With many AI tools installed you lose track of what you have and how to use it.

What I did

Claude Dashboard is the control room for my automations: a panel that shows and governs every Claude Code skill and plugin, with an always-current map of the projects. It does not call models, it orchestrates the arsenal that uses them.

The result

The whole AI ecosystem under control, from one place.

The problem

Reading a tender document of hundreds of pages to pull out hard data, frequencies and a quote is hours of work, and easy to get wrong.

What I did

Built for a facility management group. You upload the specification, even a scan, and three AIs work in a row: Claude Vision reads the pages like OCR, Claude Haiku identifies the division and structures the frequencies, Claude Sonnet writes the operational plan and fills the Excel with a risk traffic light. A few minutes later, everything is ready for the quote.

The result

From several hours to a few minutes per document, with the data ready for the quote.

The problem

Reclassifying a financial statement and computing the ratios is technical, lengthy work that few business owners can do alone.

What I did

You upload the statement in any format. The calculation (reclassification and ratios such as EBITDA, ROS, ROI, ROE, net financial position) is deterministic code, with no margin for error; Claude Sonnet steps in only where it truly adds value, to write the management commentary and flag anomalies. AI where it adds value, rules where precision matters.

The result

From the raw file to a management commentary in seconds.

The problem

In a subscription business the numbers live in separate silos: advertising, CRM, payments and banks. No single source tells you how things are really going.

What I did

A pipeline that runs on its own: n8n extracts from ads, CRM, Stripe and multi-currency banks, Postgres collects, dbt computes the metrics, Metabase shows them. A local AI model links customers across systems and consolidates cash in euros.

The result

Management sees everything at a glance, with always-current data.

The problem

Opening or changing a VAT number is a rough path: the AA9/12 form is full of technical fields and one mistake means starting over. And it is personal data to handle with care.

What I did

With VARCO you describe what you want to do and the AI triages it: type, tax regime and ATECO code, with official citations. A wizard, one question at a time, fills the AA9/12 form live, then you print the PDF. Everything encrypted in your browser.

The result

A clear path from idea to a ready form, with the data kept safe.

The problem

A showcase site is usually static. We wanted something that showed what we can actually do.

What I did

Custom, piece by piece: automations and reports in Node, a mini-CRM dashboard, leads and a daily brief on Telegram, news that updates on its own. SEO optimized for Google search.

The result

Leads, content and reports on autopilot, with no manual work.