Growth Strategist, Performance Marketer, Developer, Data Scientist and Unicorn.
Hi, I'm Georg, a London based Growth Hacker, Data Scientist, Developer, and most recently AI Engineer with 14 years of experience and a hands on approach to turning ideas into thriving businesses. I've helped startups, ecommerce stores, and SMEs achieve fast yet sustainable growth through data driven strategy and practical execution. What sets me apart is simple: I don't just talk the talk, I walk the walk. I execute directly, cutting out inefficiencies and delivering measurable results. I primarily offer my services on a commission basis, effectively covering my own cost, because that's how confident I am in what I do.
If you're building AI-powered SaaS for the DACH market, you're not dealing with one regulatory framework. You're dealing with at least five: the EU AI Act (directly applicable in Austria and Germany), GDPR (enforced differently by each country's DPA), Austria's Digital Austria Act 2.0 and KI-Servicestelle, Germany's KI-MIG implementation law and Bundesnetzagentur oversight, and Switzerland's entirely separate FADP with its own rules on AI, profiling, and personal liability. This post maps the specific nuances for Austrian, German, and Swiss businesses, shows where the regulations overlap and where they diverge, and provides the practical architecture decisions that let you ship AI features across all three markets from a single Rails codebase.
A developer's practical walkthrough of Data Processing Agreements, Privacy by Design, and Data Protection Impact Assessments for AI features. Not the legal theory, the actual Rails code and architecture decisions you need to make before shipping AI features to production. With real examples from four production Ruby on Rails applications: GrowCentric.ai (marketing optimisation), Stint.co (marketing dashboard), Regios.at (regional platform), and Auto-Prammer.at (automotive marketplace on Solidus).
When GDPR's Article 17 was written, 'erasure' meant deleting a row from a database. In 2026, it means something far more complicated. If a user's data was used to train an AI model, deleting the database record isn't enough. The data has been absorbed into model weights, influencing predictions for every subsequent user. The EDPB has made right to erasure its coordinated enforcement priority for 2025-2026, with 30 data protection authorities investigating how organisations handle deletion requests. And the Italian DPA already fined OpenAI 15 million euros for, among other things, failing to handle training data properly under GDPR. This post explains what machine unlearning is, why it's a nightmare for developers, and what practical architectural decisions you can make right now to avoid the problem in the first place.
The EU AI Act's biggest enforcement date is August 2, 2026. That's less than five months away. High-risk AI system obligations, transparency rules, and the full enforcement framework all go live on that date. If you build SaaS products that use AI and serve European customers, this directly affects you. This post explains the four risk tiers, how to figure out which one your product falls into, what the obligations actually mean in practice, and what you should be doing right now. No legal jargon. Practical guidance from someone building AI-powered SaaS for the European market.
Most marketing automation in 2026 is still glorified email scheduling. Send this email on day 3. Send a follow-up on day 7. If they click, send offer A. If not, send offer B. That is not AI. That is a flowchart. Real AI-powered marketing automation means dynamic pricing that adjusts to demand in real time, personalised product recommendations that learn from behaviour, predictive churn detection that intervenes before customers leave, and autonomous campaign optimisation that reallocates budget without waiting for a human to notice what is happening. This is what I am building with GrowCentric.ai, and this is what I implement for ecommerce clients on Rails and Solidus.
Most AI projects fail. Not because the technology is bad, but because the data is messy, the systems are old, and nobody knows where to start. Gartner predicts that 60 percent of AI projects will be abandoned due to poor data quality. This is the where do I even start post. Data quality, legacy systems, realistic first steps, and real examples from Rails, Solidus, and SaaS projects I have actually built.
