What's relevant?
Over a decade of retrieval thinking from the world of search engines. Today, I'm using that experience to understand LLMs from the ground up — and explore what modern AI systems are really made of.
The Journey
Pioneering work — again and again.
Three chapters, one pattern: curiosity about systems nobody has quite figured out yet. Self-taught, from the ground up — someone who genuinely enjoys testing, analyzing, and figuring out how things work.
SEO Analyst.
Started with data. Reports, patterns, endless spreadsheets. The kind of mind that likes to poke at things — and wants to know why one thing works and the other doesn't.
Technical SEO.
Under the hood. How search engines actually work — crawling, indexing, ranking. Back then, there weren't many playbooks for this. Everything self-taught, from the ground up. Pioneering work.
AI Research.
Same mindset, new field. LLMs instead of Googlebots — but the core question hasn't changed: how does relevance get into the system? Today, the answer is RAG & Embeddings.
Worked alongside dev teams at NewWork SE (XING SE) — a company that has rethought work from day one.
Method
Analyse › Understand › Advise
No guesswork, no gut feeling. Data first, then the picture — then the recommendation that actually fits. In that order.
Coding across multiple languages.
Hands-on experience with the languages that matter — on the web, on the backend, and everywhere in between. From quick scripts to production code.
Sparring with dev teams.
Experience from an environment that reinvented how work works — agile methods, product thinking, cross-functional teams working in sync.
"Knowing how large dev teams operate — learned inside a company that lived New Work from the very beginning."
— NewWork SE (XING SE)RAG & Embeddings.
An LLM is only as good as the context it gets. I build systems that decide which content is relevant, and when — instead of dumping everything at the model and hoping for the best.
This is essentially the same problem search engines have been solving for 25 years. Relevance scoring. Retrieval. Context engineering. The surface has changed — the principles haven't.
RAG architectures
Rule-based retrieval systems that don't leave context to chance — especially useful for coding, where data volumes are huge and selection makes the difference.
Embedding strategies
How does content get vectorized, indexed, matched? Which chunk size, which metadata, which re-ranker? Systematic testing, not gut feeling.
Model benchmarking
Different LLMs head-to-head, under real conditions. Which model delivers the best results for which task, with which context?
Context engineering
Prompt architectures, context window optimization, dynamic routing. The rules that decide what the model gets to see, and when.
The Foundation
Technical expertise, analytical thinking.
The foundation everything else is built on. Over eight years working at the intersection of search technology and data-driven performance marketing — still very much part of the day-to-day.
Technical SEO
Deep technical audits, crawl optimization, Core Web Vitals, structured data — finding what others miss.
Paid Ads & Analytics
Data-driven Google Ads management, conversion tracking architectures, and deep analytics work — turning spend into measurable performance.
Contact
Got a project in mind?
Whether it's AI consulting, SEO/GEO, or paid ads — feel free to reach out.