Most AI projects don't fail on the model. They fail on integration with existing systems, on data protection, and on teams that don't come along. I design and build AI systems that fit your workflows, keep your data in-house, and that your team can run after handover. Behind it: 15 years of software practice and a PhD in speech AI.
Currently: AI workspace pilot for a large fitness and wellness operator with several hundred thousand members.
Most enquiries don't start with “we need AI”. They start with a problem that costs time every single day.
Quotes, contracts, manuals: your team searches instead of working, and when someone leaves, the knowledge leaves with them. What I build: an assistant that answers questions from your own documents, with sources.
Private accounts, customer data in the chat window, the works council asking questions. What I build: a governed AI workspace with roles, rules, and training, GDPR-compliant and agreed with the works council.
Impressive demo, but no integration, no owner, no operations. What I build: systems designed for operations and handover from day one, integrated with your systems, documented, and carried on by your team.
Strategy where you need it, working software where it counts. For companies of 50 to 2,500 people without an in-house AI team: one point of contact from the first assessment to handover to your team.
Turn chatbots into reliable tools: assistants that know your documents, work inside your systems, and take over routine tasks, from quote to ticket.
One governed access point to AI for everyone: roles, rules, data-protection guardrails, and training instead of shadow AI on private accounts. Piloted small and built to scale.
Speech recognition and voice assistants on scientific foundations, on your own hardware if your audio can't leave the building.
From idea to running product: web, app, device, and operations from one source, without coordination loops between five trades.
Is it worth it? Buy or build? You get feasibility checks, roadmaps, and second opinions on vendor offers, from someone who implements them.
Churn, demand, quality: models built on your data, integrated into your daily operations, with measurable benefit.
Real numbers from real projects in research, enterprise, and industry.
What I learned in enterprise projects and research used worldwide, I bring to companies without an in-house AI team. The common thread is software that ships.
Sprint-based, demo-driven, and built for handover from day one. You see working software every two weeks, not slideware, and you end up with a system that keeps running without me.
We scope your problem and I give you a straight answer on feasibility, effort, and fit. Free, no obligation.
Two weeks, one clear goal, working software demoed at the end, documented in English or German.
Continue sprint by sprint as long as it pays off. Either way, the code, docs, and know-how are yours, your team or your IT provider can run the system, and I stay reachable if you want.
AI that runs. Teams that come along.
Technology is half the job. For change, training, and internal communication, as well as data protection, works council, and the EU AI Act, I work with experienced partners. You get a team, not a lone wolf.
I've spent 15+ years building software and a good decade of that building AI. From a PhD in speech AI at LMU Munich (magna cum laude, 14 published papers) to leading data science teams in consulting and leading enterprise-wide projects at Wacker Chemie AG.
Today I do what I like best: working directly with teams to take AI from idea to adopted product. Programming and translating between business and technology included.
Under the hood