Data and knowledge graph architect. I turn messy enterprise data — multi-source, poorly structured or legacy — into reliable, usable models, ready for analytics and artificial intelligence.
27 years building data-driven business platforms. I master the whole chain: understanding the business, modeling the data, designing the architecture and coding the system. I make data clean, structured and intelligible enough for analytics and AI to finally become useful.
My signature: a single object / attribute / relation meta-model, built and refined for 25 years — knowledge graphs before the term existed.
In plain terms: a classic database tells you what you own; a knowledge graph knows how everything is connected — and answers questions a table cannot handle, like “if this supplier fails, which products are affected?”.
I work with SMEs, manufacturers, product groups and organizations whose business data is rich, historical and scattered — across Excel, ERPs, internal databases, documents and line-of-business tools.
From a quick diagnostic to advanced exploitation — most start with a short, scoped engagement. Do you recognize one of these situations?
First, know what your data really says — and where it betrays you.Deliverable: a map of your data, its inconsistencies and quick wins.
What your data truly allows, and what holds it back.
Detect late, batched or after-the-fact entries that skew your dashboards.
Clean, connect, migrate: turn a scattered data estate into a reliable repository.Deliverable: a consolidated, reliable, ready-to-use repository.
Tell apart the different realities hidden under a single label.
Own the gap between the old system and SAP, and catch unworkable rules before go-live.
Structure scattered documents and histories into a reliable corpus an AI assistant can use.
Once the data is solid, turn it into decisions: trends, signals, predictions, memory.Deliverable: indicators, signals and first usable ML/AI models.
Neutralize structural effects to see the real drifts, invisible in averages.
Separate the seasonal cycle from the real anomaly, and focus prevention at the right time.
Cross reviews, actual tasks and career paths to see a departure brewing.
Cross operations with finance to price what nags you.
Trace decisions and effects: “in a similar case, here is what was decided, and the result”.
The outcome always depends on your data and your context. What I guarantee is not a number promised in advance — it is a rigorous method to look for it, and the honesty to tell you when a lead goes nowhere.
Most engagements begin with a short diagnostic: a review of your sources, inconsistencies, hidden rules, migration risks and AI potential.
Deliverable: a map of your data, priorities, quick wins and an architecture recommendation.
No prefab solution, no technical jargon. A simple approach you can follow end to end.
I start from your business problem and your data as it is — not from a theoretical model imposed from outside.
I fit your data into as few rules as possible. Each exception that resists reveals a business rule or a forgotten piece of history — and the real model emerges.
What works and what doesn't. You always know where we stand, and why.
The result lives in your tools and keeps serving long after I am gone.
Three proofs, three qualities: maturity, robustness, pedigree.
Qibud is the strategy-execution steering platform I design and keep evolving. A temporal knowledge graph that embeds machine learning and AI — the same object / attribute / relation model I have refined for 25 years, taken to its full expression. It is where I keep my craft at the frontier, and it is that rigor I bring to your data.
Used by demanding organizations, from luxury jewelry to construction.
My client mandates stay the priority, delivered with a clear commitment to availability and deliverables. Qibud is my lab, not a scheduling conflict.
Qibud is not a prerequisite: I work directly within your existing systems, tools and data.
Strategy-execution platform built on a universal object / attribute / relation model. Temporal knowledge graph, embeddings, ML/AI indicators as time series.
Processing and structuring of high-volume business data; ERP-to-SAP migration.
Platform linking patents and products, born from EPFL research. Large-scale crawling, parsing and machine learning, funded notably by the Sloan Foundation.
Back office for the cantonal vehicle registration services. An extranet designed back then is still running today.
Designer of Centrix (PIM) for luxury and industry: Swatch Group, LVMH, Richemont, Tornos. Graph architecture, automated imports of complex data. Team of ~12 engineers.
“Orange Mate”: automated generation of a 300-page weekly report, replacing an insecure Excel file.
12th employee. Extranet framework for the International Olympic Committee; the mINTEGRATE product showcased at the Orbit trade fair in Basel.
At eleven, on an Amstrad, I was programming hockey league simulations: I weighted the teams, ran standings and cup brackets. I was already handling data without knowing its name. Thirty-five years later, I do the same thing at the scale of major industrial groups.
For a long time, I thought I had held several jobs. In reality, I refined a single idea: a model where everything is object, attribute and relation, able to absorb business complexity without locking the company into a rigid architecture.
What I love: receiving “messy” data, 80% correct, and fitting the rest into as few rules as possible. Each rule reveals a business exception or a forgotten piece of history. When everything clicks into place, I know the model is right — and that the data can finally speak.
Available for contract work in Switzerland and remotely.