Data-platform architecture
Architecture, roadmap, QA and acceptance of institutional data platforms — data model, governance, security-by-design.
Independent principal · Technical & data-architecture direction
I turn complex data problems into governed platforms — and the team that ships them.
A bounded engagement, principal direction: from an ambiguous requirement to a governed architecture — delivered by your team, which I enable, or mine. 25 years in public-data systems; since 2024, the AI layer on top.
01
What I direct
P1
P2
AI over data
Idea to production: LLM monitoring in production, natural-language access to data (SQL, SPARQL), RAG over documents, MCP integrations.
P3
Agentic method
How I work and how the team I enabled works — architecture, analysis, ops, delivery. Not advice from outside.
P4
Command of the stack
Depth to the metal as the source of authority to direct. I set the road and justify every call in full.
02
How I work
The two questions everyone actually has — do you manage people? do you deliver, or just advise? — answered plainly.
Direction, not line-management
I own architecture, roadmap, the quality and security bars, and release acceptance. I don't carry your headcount, and I'm not your project manager.
The team is yours, or it's mine
I direct and enable the team you already have — or I bring the delivery team: DEPP, the firm I co-founded. The answer to "who ships it?" is explicit.
A bounded engagement
Defined scope, on personal P.IVA: from ambiguous requirements to architecture, infrastructure, and a team on a clear path to ship.
03
Selected work
01
MAPS data-lake
Directing an institutional data-lake — technical design, QA Playbook, security-by-design, release acceptance.
Gran Sasso Tech · 2025 → 02LLM monitoring
Categorizes, summarizes and prioritizes the news stream against each client's interests. In production.
Public-affairs · 2024 → 03OpenCoesione.gov.it
The data architecture behind Italy's national portal for EU cohesion funds. Built to be queried and trusted.
National portal 04Health-authority system
From ambiguous requirements to architecture, infrastructure and a development team on a clear path to ship.
Custom system
About
From the mathematics, not the hype cycle
- BasePersonal P.IVA · remote
- LanguagesEN · IT
- BackgroundOpenpolis · DEPP · GST
- OriginPhysics / neural networks, 1996
I'm an independent principal architect. For 25 years I've turned complex public data into platforms institutions rely on — from OpenCoesione's national data architecture to the MAPS territorial data-lake I direct today.
I built data infrastructure for public accountability: OPDM, the platform behind Openpolis' monitoring of politics and institutions; the national archive of CIPE deliberations; the Popolo international standard for data on people and organizations. I'm one of the people who built Openpolis into a national foundation, and I co-founded DEPP, where I'm CTO.
Since 2024 I design the AI applications that put data to work. The lesson, simpler than the hype: an AI application over data is only as good as the platform beneath it — I design both layers.
Contact
A data-heavy problem that needs direction?
Bring me the ambiguous data problem: I turn it into a governed platform and the team that ships it — yours or mine. Architecture direction · AI-application design · technical direction. Remote-first, in English or Italian.
guglielmo.celata@protonmail.com →