Davide Scuteri Moretti
I design the computational institutions through which data become explainable, governable and reusable knowledge.
Senior Data Architect and technology leader specialising in the design of governed Data Platforms, semantic Lakehouse architectures and computational systems for healthcare, scientific research, emergency response and enterprise operations.
Senior Data Architect & Head of Development for Data
- —Defining logical, semantic, application and technology architectures.
- —Coordinating the evolution of OCTAVIA as a computational Data Framework.
- —Supervising Data Object modelling and Bronze, Silver, Gold and Data Product boundaries.
- —Establishing ingestion, transformation, reconciliation and publication patterns.
- —Designing metadata-driven and contract-first pipelines.
- —Coordinating integration across relational, semi-structured, streaming, document and scientific sources.
- —Specifying data-quality, lineage, quarantine and replay mechanisms.
- —Defining privacy, masking, pseudonymisation, access and retention controls.
- —Guiding engineering teams from architectural design to executable pipelines.
- —Aligning platform architectures with advanced analytics, machine learning, process mining, graph analysis and generative AI.
- —Translating business, clinical, scientific and operational requirements into implementable technical specifications.
Eight interlocking commitments that define how I design Data Platforms.
Data Objects as Semantic Contracts
A Data Object is a governed computational entity with its own identity, grain, temporal semantics, source evidence, quality obligations, privacy classification, lifecycle, schema version, lineage and publication contract. A single source row may contribute to several Data Objects; a single enterprise object may be reconstructed from several source records.
Task-Based Lineage
Lineage is formalised through source columns C, computational tasks T and Data Objects D: f₁: C → T, f₂: T → D, g = f₂ ∘ f₁. The composition provides a formal path from each source field to the governed object in which its meaning is materialised. Tasks are arranged in a Directed Acyclic Graph.
Evidence Before Interpretation
Raw evidence is retained as an auditable source. Normalised or analytical representations never silently replace the original payload. Append-only ingestion, record and payload hashes, schema and calculation versioning, quarantine without source loss, idempotent replay and explicit matching decisions are constitutive.
Data Quality as Executable Architecture
Quality is encoded into the computation through parsing rules, datatype and domain validation, referential checks, numerical reconciliation, completeness thresholds, cross-layer record-count controls, controlled reject tables and publication gates.
AI Readiness by Design
Feature provenance, training-set boundaries, model inputs, source evidence, confidence scores, text embeddings, graph features and derived calculations remain connected to the lineage that produced them.
Privacy, Masking and Pseudonymisation
Direct identifiers are separated from working representations through pseudonymised keys. Masking, access, retention and audit are declared in the same contract space as identity and quality clauses, and are evaluated by the same runtime.
Selective Quarantine and Replay
Failure without total arrest. Non-conforming records are quarantined with full evidence; publication gates are respected; corrective replay is deterministic and idempotent.
Contract-First Publication
Publication is governed by the same Data Contract used for validation, with explicit temporal, structural, semantic and governance clauses; consumers depend on the contract rather than on the pipeline.