Building AI products and the
teams behind them.
15+ years turning research and machine learning into products people actually use. This is my corner of the web: what I'm building, what I'm reading, and what I'm still figuring out.
Currently Head of AI Platform & Agents Foundry at Conrad Electronic SE
From the blog.
Notes on AI in practice — papers worth reading, tools worth using, and lessons from building and leading AI teams.
What I build.
AI search & a multi-agent sales engine
Hands-on ownership of the architecture, implementation, and quality gates that keep agentic systems reliable in production - plus the roadmap and stakeholder communication for the platform behind AI-powered search and a multi-agent sales engine.
Recommendations & fraud detection at scale
The cross-functional teams I direct built the recommendation engine that powers the online shop, plus a fraud and order-risk system now relied on daily across sales and customer care.
LLM automation across support & operations
LLM-based solutions to handle Zendesk tickets routing and automation of the process of Request for Quotation(RFQ) alongside Google Document AI + Steering the automation for the process of product categorization and supplier onboarding using in-house language model.
Fraudulent-listings detection
A rules-based process gave way to a scalable, end-to-end ML product that sharply cut manual listing checks, backed by computer-vision models for image quality and duplicate detection.
Certificates & badges.
Verified badges and course certificates I keep up to date — from Credly, Coursera (including Google Cloud), Udemy and more.
How I got here.
My path started in research: several years at Inria and Télécom ParisTech on large-scale image retrieval — hashing methods to approximate SVMs at scale, boosting for interactive object retrieval, and computer-vision pipelines for automated plant species identification, work that led to a handful of peer-reviewed papers. That period is also where I built my long relationship with C++ and Python, in the years before AI pair programmers existed — every prototype meant writing and debugging the algorithm yourself, and understanding it well enough to make it both fast and correct. That discipline has stuck with me since. Eventually I wanted to see ideas ship, so I moved into engineering — building data products and ML systems across Munich's startup scene, then senior machine-learning work at Scout24.
For the past several years I've been building and leading data and AI at Conrad Electronic — from a single-hire founding of the data science function, to Head of Data Science, to Head of AI & Data Platform today, with 5+ years now in people leadership. I help shape the AI strategy and drive its execution, staying hands-on on system architecture, MLOps/LLMOps, and cloud platforms (GCP, AWS) for high-impact, cross-functional work. I've also built the team from the ground up — recruiting, mentoring, and investing in people's growth as the function scaled. I lead as a servant leader, and I stay close enough to the work to be accountable for what ships.
Outside the day job, I founded PyData Munich and still write here about doing this work well.
Building the room, not just the product.
PyData Munich Meetup — founded in 2016 and organized ever since, bringing together the city's data and AI practitioners. Earlier, a long-time open-source advocate: founder of an OSS club and organizer of events and conferences.
Always glad to trade notes with fellow data & AI practitioners, explore a collaboration, or talk about building teams that ship. Drop me a note below or find me on LinkedIn.