I own the manufacturing decision loop — detect, measure, explain, intervene, standardize — turning fragmented plant, lab, supplier and ERP data into causal root causes and measured shop-floor change.
I turn fragmented plant, lab, supplier & ERP data into causal root causes, production policies, and measurement gates operators actually use.
Not just models — measured change in scrap, cost and process behaviour, in production. Detect a problem, measure it objectively, explain why it happens, intervene at the right lever, and standardize the fix so it sticks.
My edge is the rare overlap most heavy-industry data work is missing: foundry metallurgy, ERP / SAP master data, and data science in one head — the moat that turns a model into a trusted shop-floor decision. 14+ years across automotive, marine and aerospace casting.
And I've run the floor, not just analysed it — Technical Director / Head of Foundry leading operations and capital projects. Two tracks, one engineer: production & operations leadership and manufacturing data science.
Integrated five data systems across ~2,000 production charges. Causal inference + Design of Experiments isolated a supplier-chemistry × furnace-state interaction — driving a zero-CAPEX process-policy change plus a production fix, and replacing subjective inspection with quantitative acceptance criteria.
Modernized the foundry's engineering: introduced a simulation-first casting-design approach (Flow3D) and data-based process development, and streamlined prototyping to sub-5-day 3D-printed-mould iterations. The new method lifted yield +50% and cut chill usage −60%, and root-caused magnesium & copper-alloy defects — incl. design support to Rimac.
Production computer-vision — fine-tuned instance segmentation for automated grain sizing. Offline desktop app with an active-learning correction flywheel; now the objective measurement gate.
An MCP server making auditable analysis the path of least resistance: enforced pre-registration, hash-chained lineage, deterministic outputs. Proven end-to-end on a real foundry root-cause study.
Analytical data cube + relationship graph + automated 10-pillar governance audit + SPC — built to expose cost and process deviations invisible across siloed systems.
~90 models across five defect types with SHAP-driven process-driver science and rigorous feature engineering — scrap moved from after-the-fact reporting to forward forecasting.
A personal "second brain" — an MCP server + desktop app that turns a knowledge vault into an agentic system: semantic search, schedulers, and multi-model orchestration the AI calls to run real workflows end to end.
A standalone app that automates foundry scrap-rate roll-downs and SAP write-back from the unified data cube — replacing manual corrections with a governed, repeatable pipeline.
A personal software framework — frozen interfaces and inherited test suites, so new projects don't reinvent storage, auth, or scheduling from scratch. The discipline is verification before convenience: nothing ships until an independent check agrees it works, not just the person who wrote it.
Foundry Manager (2022–2024), then realigned to the data & analysis expert role. Owns the manufacturing decision loop — the Closed-Loop Quality Intelligence program (root-caused a ~CHF 5M scrap defect open 5–6 years) plus the production-intelligence stack (CERES, Vulcan, BIOME, V3CTOR). Plant-wide data management; 5S, Kanban (~30% lead-time gain), automated core-shop vision.
Modernized the foundry's engineering — simulation-first casting design (Flow3D) and data-based process development; streamlined prototyping to sub-5-day 3D-printed moulds (+50% yield, −60% chills). Root-caused magnesium & copper-alloy defects; design support to Rimac on next-gen electric engines.
Rescued an engine-block launch by root-causing hot tearing with process-parameter changes alone. Built a predictive material-property tool from 10,000+ tensile bars to secure SOP; created an SQL framework to leverage process data across the chain.
Supported a 300,000+ t/yr foundry network; led the most demanding large castings (10-tonne wind-turbine hubs, marine engine components for clients incl. Wärtsilä); cut scrap 50% via simulation + process-data analysis.
The metallurgy & manufacturing foundation behind the data work. Lean Six Sigma Green Belt (2026).
Open to senior roles in manufacturing data science and production / operations leadership — and consulting — where metallurgy, quality and data meet. If you have scrap, cost or process problems that resist the usual dashboards, let's talk.