25+ Years in Business.
Now With AI.

25+

Years in business

9 → 1.5h

Weekly admin, after AI took over

4×

Content output tripled, half the effort

I build. I don't just advise. Every system below runs in my own business, under real cost pressure, on real processes.

Since 2002, mindmelt has been my own agency in the Rhine-Main region. Schott, Zeiss and Helaba have trusted us for years, on work where precision mattered and mistakes cost real money.

The four projects below are built in-house and running every day. Back office, content, SEO, ticketing - digital automation that actually works, shipped from genuine need rather than theory.

The Challenge

Until 2024, mindmelt ran on the usual silos. Proposals in Word. Invoices in DATEV. Customer data scattered across spreadsheets. Comms split across three channels. Admin ate eight to ten hours a week.

What We Built

A custom back-office web app. Invoices numbered automatically. Receipts generated on the fly. One central customer database. ZUGFeRD-compliant PDFs drop straight into DATEV or any modern accounting tool. Brief to finished document, no manual steps.

Results

Admin: 9 hours a week down to 1.5. Invoicing: 45 minutes down to under 5. Error rate: zero. Live since March 2025, not a single manual correction.

Stack

• Python / Flask backend with REST API
• ZUGFeRD/Factur-X for machine-readable invoices
• LibreOffice Headless for PDF generation
• Central customer DB with automatic matching

The Challenge

Content marketing was manual, end to end. Topic research, briefing, writing, SEO polish, formatting for blog and social, sourcing images. One blog post with visuals chewed up four to six hours of focused work.

What We Built

A fully automated content pipeline, built as a web app. Feed it a topic brief and it produces SEO-optimised articles (1,200-1,800 words), AI-generated visuals for each section, automated layout and direct publishing into REDAXO CMS. Approval gates at every stage.

Results

Production: 5 hours down to under 20 minutes of active work. Output tripled. Tone and SEO consistency sharpened. Every article ships with professional AI visuals in two distinct styles.

Tech Stack

• Python / Flask for pipeline control
• Claude AI for text generation and keyword clustering
• fal.ai FLUX Pro for AI image generation
• REDAXO CMS integration with automatic module mapping

The Challenge

SEO reporting across multiple domains meant pulling data by hand from Google Search Console, rank trackers and the CMS. Monthly reports took three to four hours. Data was always stale. Action plans ran on gut feel, not numbers.

What We Built

A bespoke SEO dashboard, built as a web app, with ten specialist modules: rankings, keyword research, backlink analysis, broken link detection, on-page audit, SERP features, competitor analysis, local SEO, Google Reviews and Search Console. It auto-detects REDAXO installations and fixes broken links straight in the database.

Results

Monthly reporting: 3.5 hours down to 15 minutes of review. Data freshness: daily, not monthly. AI-driven action plans ranked by impact and effort. 24-hour caching keeps API costs low without letting data go stale.

Tech Stack

• PHP 8.3 with custom routing and PDO
• DataForSEO API for rankings, keywords and backlinks
• Google Search Console API (Service Account / JWT)
• MySQL for historical snapshots and caching

The Challenge

Requests came in via email, phone and chat, all in different shapes with no common structure. Briefs had to be compiled by hand. Images were made separately. Content went into the CMS manually. Social posts needed another round of email approvals.

What We Built

An email-driven ticket system that captures requests automatically, assigns ticket IDs and processes them with AI. Hooked up to fal.ai FLUX Pro for on-demand image generation. Direct REDAXO integration pushes text and images straight into the right page structure. Approved social posts publish from the system itself.

The Results

Request to finished CMS entry: two hours down to under ten minutes. Image generation: professional output across six workflow variants (text-to-image, image-to-image, 4K, video). Every step logged with full conversation history per ticket. Error handling and status updates run automatically by email.

Tech Stack

• Python daemon with IMAP polling and SMTP responses
• Claude AI for request processing and content creation
• fal.ai FLUX Pro for AI image generation across six modes
• REDAXO CMS integration with media pool connection
• JSON-based ticket tracking with threading