Consulting & architecture

From critical data to governed AI

I design critical information systems where you can prove what happened: which data, which process, when, by whom. I extend that same demand for proof to systems that rely on AI — integrity and auditability in the architecture from the first line, never a mere end-of-project report.

My ground: platforms where a mistake shows

Banking, insurance, public finance: environments where wrong or missing data has a real cost — regulatory, financial, operational. That is where I have worked for 20 years.

My approach is constant: make the system auditable by design. We do not bolt traceability on afterwards, we model it with the flows. Control is not a step, it is a property.

That demand doesn't stop at databases. As AI enters production pipelines, the question becomes: can you prove, trace and govern what an AI-assisted system decides and produces? It is the natural extension of my craft — and the heart of my R&D.

Areas of work

What I work on

Critical systems architecture

Critical data platform architecture
Strongly auditable systems
Complex flow modelling
Distributed architectures, monitoring and flow control

Data governance & integrity

Processing traceability
End-to-end data lineage
Control and verification mechanisms
Technical governance frameworks for data flows

Data platform transformation

ETL / ELT modernisation
Platform migration (e.g. Oracle → PostgreSQL, ERP → Workday)
Pipeline industrialisation
Decision indicator standardisation

Technical delivery

Architecture scoping
Multi-team coordination (business, IT, operations)
Agile / SAFe delivery at scale
Value-driven steering and KPIs

How we work together

Scoped missions, concrete deliverables

Formats

Time-and-materials mission (long-term), fixed-price architecture scoping, or one-off expertise (audit, second opinion, technical arbitration).

Deliverables

Target architecture dossier, flow and traceability model, control mechanisms, migration plan and scripts, steering indicators.

Duration & terms

From a few days (scoping/audit) to several months (transformation, migration). Paris area or full remote. Billed via PHYDYA (SARL).

Cases — problem → intervention → result

What it looks like on the ground

Anonymised missions. Named references on request.

Agri-food cooperative

Oracle → PostgreSQL migration of a critical business system

Problem

A legacy business application (the whole shipping chain: orders, invoicing, logistics, traceability) on ageing Oracle — ~342 tables, technical debt and licensing, strong traceability and integrity-control requirements.

Intervention

A 3-phase migration architecture (Ora2Pg → staging → TypeORM transformations → target schema), end-to-end hash validation and referential-integrity constraint checks, multi-environment industrialisation (Ansible, dev → prod).

Result

A tooled, reproducible and verifiable migration — integrity proven table by table, with no loss or drift.

Insurance & assistance Ongoing

Finance data migration to Workday

Problem

Migrating finance data from a legacy ERP to Workday, with an integrator, across many domains (parties, banking, fixed assets, invoices, warranties, claims…) and several runs — strict reconciliation required, "zero difference per stream".

Intervention

Data Migration role: scoping the perimeter object by object, migration strategy (pivots, runs, tenants), canonical business rules, accounting reconciliation, tracking and decision mechanisms (supported by AI-assisted steering with deterministic controls).

Result

A structured, governed migration, perimeters agreed object by object, on track for a controlled go-live.

Large insurance group

Directing a Data & BI service centre

Problem

Steering and modernising the group's Data & BI activities — ageing ETL chains, heterogeneous decision indicators.

Intervention

Directing a service centre (20 people, a €12M portfolio over 4 years), ETL modernisation, indicator standardisation, technical governance, value-driven steering and KPIs.

Result

Industrialised, standardised Data/BI activities, value-driven steering.

New ground

AI governance & assisted delivery

AI is entering production pipelines: the same demand for proof applies. I design auditable AI architectures — governed agents, quality gates, decision traceability, formal verification of invariants — so that what an AI-assisted system decides and produces stays verifiable end to end.

See my /gov framework

Auditable AI architectures & decision traceability
Governed multi-agent workflows (quality gates, cross-validation)
Formal verification of critical invariants

Method

How I work

Scope before building

I start by understanding the business stakes and real constraints — not by picking a technology. The target architecture follows the need for proof, not fashion.

Auditable by design

Traceability, lineage and control mechanisms are modelled with the flows. Every process leaves a verifiable trace. Nothing is added "at the end".

Industrialise and hand over

Tooled pipelines, standardised indicators, coordinated teams. I deliver a system your teams can run and evolve without me.

Technical environments

What I handle day to day

Data & platforms

AI & orchestration

Verification & proof

Cloud & delivery

A critical system to design or harden?

Migration, compliance, control mechanism, architecture scoping: tell me the context.

Discuss your architecture