AI SOLUTIONS · DATA · TECHNICAL LEADERSHIP
AI Solution Architect & Engineer
Principal Data Architect
Technical Leadership
Freienbach, Switzerland
I turn complex AI and data problems into working systems.
I design, build and operate production-grade AI applications and business-critical data systems, connecting business requirements, enterprise applications, APIs and reliable data pipelines. I combine original thinking with research, experimentation and hands-on engineering.
I build systems that last, and I mentor engineers, sharing knowledge they keep using.
SELECTED WORK
Swisscom
Architected and built the SIMS AI Development Kit to connect plain-language objectives with enterprise knowledge and tools, taking work through specification, implementation, testing and review. Task-specific context, multi-agent orchestration and cross-vendor LLM-as-judge evaluation help uncover omissions and refine solutions. Model benchmarks, cost and latency monitoring, automated checks and human approval controls guide delivery.
SWISS / Lufthansa Group
Architected the highly available, real-time booking-data platform in 2015, now used by every Amadeus-based airline in the Lufthansa Group and supporting hundreds of interfaces and APIs. For the Group rollout, purpose-built change detection reduced its storage footprint by around 90% and CPU consumption by half.
Simple
Combined AI and data engineering with solution design and technical presales. Built LCA Loader: 136 datasets across 63 emission-factor sources and 4.2 million profiled rows in a traceable factor pool. Implemented reusable feature-data storage and retrieval for the AI pipeline and contributed to monitoring LLM cost, latency and behaviour.
WWF Germany / WWF Kenya
Built an auditable research pipeline assessing and ranking 1,059 crop and tree species across 126 signals, combining scientific sources and AI enrichment with climate and invasiveness constraints for semi-arid landscapes. The work includes monitoring LLM cost, latency and behaviour.
Digitec Galaxus / Migros
Built and operated Central Query Store to make SQL performance and attributable Azure costs visible by query and team, turning specialist investigation into self-service tooling. Advised around 350 developers and contributed to AI adoption through the AI guild.
Independent R&D
Built a research-to-implementation workflow connecting papers, engineering specifications and experiments with a real-time data platform. Implemented reusable ML feature-data storage and retrieval, with Polars, DuckDB, Parquet and GPU processing. Model training, tuning and testing are ongoing.
WHAT I BRING
I connect LLM applications to enterprise knowledge, APIs and deterministic tools, with task-specific context, multi-agent orchestration and cross-vendor evaluation. The work spans research, specifications, software, testing and document review. Reusable knowledge, regression checks, quality and security gates, automatic fallback and human review support reliable delivery. Cost, latency and behaviour monitoring inform model and workflow choices.
Deep SQL Server and Teradata expertise: execution plans, indexing, partitioning, high availability and point-in-time history. I build real-time operational data systems and cross-technology lineage across languages, libraries and storage systems, with awareness of data sizing, indexing and partitioning. Performance investigation and query-level cost attribution become tools engineers can use themselves.
Reusable feature-data storage, retrieval and population pipelines connect AI applications and ML models to reliable inputs. My work spans streaming ingestion, data contracts, schema and drift controls, Polars, DuckDB, Parquet and GPU processing. Independent time-series research links papers to specifications and experiments, with evaluation covering temporal leakage, feature ablation, uncertainty calibration and CPU/GPU consistency.
I translate business requirements into implementable architectures, align decisions with stakeholders and stay hands-on from customer discovery and business case through delivery and operation. I mentor engineers and weigh performance, cost, security and risk. Business ownership, technical due diligence, M&A and an ISP turnaround inform my judgement; at eTel, I headed half of a 160-person technical organisation.
QUALIFICATIONS
Dipl.-Ing. in Computer Science, TU Wien, 1997. MSc-equivalent degree, graduated with distinction.
Academic focus on neural networks, computer vision and business IT.
Continued study in deep learning, time-series ML, reinforcement learning, LLM alignment and advanced agentic systems, including EPFL Applied Machine Learning Days workshops.
Python, C#/.NET, Rust and PowerShell; SQL Server, Teradata, PostgreSQL, Polars, DuckDB and Parquet. Code knowledge graphs and data lineage connect dependencies across languages, libraries and storage systems.
Azure SQL workloads and cost analytics; AWS migration preparation; Google Cloud project environments; Docker/Rancher deployment and operation. AI integration through CLI, APIs, VS Code and MCP; applied ML work spans research, implementation, training, tuning and testing.
German: native. English: C2 and French: B1, self-assessed.
Experience across airlines, telecommunications, retail, financial services, sustainability and broadcasting, alongside independent quantitative-finance R&D.
LET’S CONNECT