Freelance engineering from Hamburg, for technical leads and teams that want calm senior help.
I work end to end: identifying where AI creates real value — for users, or inside slow and expensive internal processes — and shipping the systems that make it real. Calm planning, clear implementation, systems that hold up after I leave.
7+ years of production work · 13 client projects delivered · 8+ systems in production. Work across Mercedes-Benz, Vodafone, Hapag-Lloyd, Merck, and Deutsche Bahn.
Trusted inside teams at
Selected places where the work had to be understandable, reliable, and useful after handover.
Selected work
Most useful work is not a shiny demo. It is the boring-good layer underneath: permissions, deployment paths, data flow, observability, handover, and enough architecture that the next team can keep moving.
Mercedes-Benz Tech Innovation
Designed and implemented a scalable backend architecture and agent gateway enabling internal teams to develop generative AI and agent-based applications.
Stack: AWS Lambda, API Gateway, DynamoDB, Terraform, Step Functions, SageMaker.
Vodafone
Designed and implemented a scalable ML platform architecture to standardize model deployment and lifecycle management across teams.
Stack
AWS SageMaker, Airflow, DVC, Lambda, API Gateway, S3
Proof
Standardized ML deployment workflows Reduced operational overhead for data science teams
Hapag-Lloyd
Built and deployed NLP models automating email classification in global logistics workflows.
Stack
BERT, AWS Lambda, ECS, API Gateway, DynamoDB
Proof
Processes over 190,000 emails per month automatically Winner of WirtschaftsWoche Best of Consulting Award
How I help
I am happiest close to real constraints: existing teams, production workloads, unclear architecture edges, and features that need to become maintainable systems.
Design and implementation of scalable, secure backend systems and cloud infrastructures for modern applications.
Typical work: Serverless architectures (AWS), Infrastructure as Code (Terraform / CDK), Microservices & API architectures, CI/CD and automated deployments, Secure cloud networking & IAM.
Development of production-ready AI applications including LLM systems, intelligent agents, and AI-driven automation.
Typical work: LLM-powered applications, Agent architectures, Retrieval-Augmented Generation systems, AI automation workflows.
Development of modern web applications, SaaS platforms, and scalable backend systems.
Typical work: NextJS web applications, SaaS dashboards and analytics tools, Authentication systems, API-driven backends, Landing pages and product interfaces.
Helping organizations evaluate, design, and implement AI initiatives.
Typical work: AI opportunity discovery, Use-case evaluation, AI system architecture design, Architecture reviews, AI governance & risk strategy.
Side projects
This is the less formal proof: product instincts, curiosity, and the habit of finishing. Some are public, some are private, all of them taught me something.
QR-based digital menu platform helping restaurants replace physical menus.
Benefits: No printing costs, Real-time updates, Mobile-friendly menus.
Serverless backend architecture for a generative AI mobile application.
Focus: LLM integrations, Serverless infrastructure, AI-powered mobile experiences.
Interactive tax calculator web app.
Demonstrates: Modern web UI, Data-driven applications, NextJS frontend architecture.
Private operator dashboard for monitoring a VXX-focused trading bot.
Highlights: Supabase-authenticated operator UI, Realtime trading state surfaces, Path-integrated deployment at /apps/tradingbot.
Working notes
Friendly, direct, useful. I like clear trade-offs, small artifacts people can keep using, and implementation that does not require a ceremonial reveal.
Understand goals, constraints, and success criteria.
Define a pragmatic architecture aligned with project scope, from lightweight prototypes to scalable production systems.
Build reliable systems using cloud-native architectures and automation.
Production deployment, documentation, and support for internal teams.
Contact
A few rough paragraphs are enough: what you are building, what feels uncertain, and what needs to be reliable.
I usually respond within 24 hours.
No form, no funnel. Just a normal email.
