Carlo Cancellieri
I design and deliver geospatial platforms and distributed data systems, combining hands-on software engineering with architecture, infrastructure as code and the software development life cycle (SDLC) across cloud and on-premises environments. Open-source work and international collaboration shape how I build, release and operate them.
Explore by the decisions and disciplines that interest you. Each path connects projects across a broader engineering career.
All paths and projects are shown.
Boundaries, contracts and decisions that make systems easier to evolve.
Geospatial engineering and information management through open standards and reusable access to spatial data.
Infrastructure as code, SDLC and operations across cloud and on-premises engineering.
Identity and access management, trust boundaries and delivery controls that support DevSecOps.
Engineering judgment, collaboration and knowledge transfer, with an incident-management approach grounded in diagnosis and recoverable operations.
Responsible AI, research and the decisions that connect technology to business.
My engineering roots are in Linux, native programming and distributed data. A computer science thesis on grid environments led to the C++ SQL Handler for OPeNDAP Hyrax and Google Summer of Code. At GeoSolutions, I contributed to GeoServer and learned to build within an open-source community.
Today, at FAO, my work spans platform architecture, implementation and operations: GeoID and DynaStore, data discovery, identity, cloud infrastructure and the workflows that connect them. The work happens with engineers and colleagues across teams and disciplines.
I move between Linux, macOS and Windows, and between languages as the problem requires. Architecture matters to me when it helps a team make decisions, operate a system and change it with confidence.
Institutional platforms, community infrastructure and independent experiments. Each project describes my contribution; technical detail and source evidence remain available below.
Read the problem, architecture, decisions, outcomes and learning behind the work.
I developed the original JMS clustering plugin and related articles at GeoSolutions. It propagates catalog and configuration changes between GeoServer instances while multiple nodes serve queries. Spatial datasets have a separate sharing or synchronization strategy.
The original development was committed in SVN and later migrated to Git. The current module includes subsequent community maintenance.
I developed ckanext-jsonschema for structured metadata forms, validation and Solr indexing. Separately, I worked on an Eventarc-based indexing design across Elasticsearch, Solr and object storage. The event-driven integration is separate from the public extension.
My work combines API extensions, domain modules, operation-specific storage routing and recoverable background tasks. A DuckDB case study queries supported geospatial files in place, while external search indexes are maintained as asynchronous projections.
DynaStore and its institutional deployments are FAO work delivered with colleagues and collaborators.
FAO's modular, multi-tenant geospatial platform for turning OGC API and STAC standards into deployable services. Its extension architecture covers discovery, vector and raster access, maps and tiles, processing, joins, dimensions and working 3D delivery paths.
GeoID remains an official FAO initiative. This portfolio describes my technical contribution; project ownership and institutional inquiries belong to FAO.
A multi-tenant geospatial engine developed at FAO for catalog, map, tile and vector-data workflows. Its architecture separates protocol handling from storage and operational services, with pluggable PostgreSQL/PostGIS, search and analytical data paths across cloud and on-premises deployments.
My contribution covered the architecture and technical leadership of these shared services; DynaStore and its institutional deployments remain FAO work.
Since 2023, I have contributed to the DynaStore geospatial catalog program. Its architecture separates API extensions, service modules, storage capabilities, and background tasks, with external search indexes maintained as asynchronous projections. A related DuckDB case study queries supported geospatial files without first copying them into a database.
Engineering for FAO's corporate statistics platform: frontend modernization, statistical processing and repeatable delivery within an international team.
My contribution included React/Redux/TypeScript frontend work, R-based processing pipelines, and CI/CD and software delivery practices for the SWS team. The platform supports Food Balance Sheets, Supply Utilization Accounts and domain-specific statistical methodologies.
FAOSTAT public repositoriesCore contributor to GeoServer, GeoNetwork, GeoTools, and GeoBatch. Led development of advanced features and performance optimization tools.
At GeoSolutions (2010–2014), led development of enterprise features for OGC Web Services, metadata management, and high-performance geospatial data serving.
I developed the original GeoServer JMS clustering plugin as its sole developer from 2010 to 2014. It synchronizes GeoServer configuration changes between instances through JMS. The current module also includes later maintenance. My related article explains the clustering approach.
The original development was committed in SVN and later migrated to Git; source attribution preserves the historical authorship.
I contributed to the development and maintenance of GeoServer-manager, a Java client for the GeoServer REST administration API. Multiple contributors worked on the project, and its public repository is archived.
GeoServerStressTest is a supporting tool for exercising GeoServer under load and evaluating service behavior. It complements my broader GeoServer ecosystem work.
A C++ BES module enabling relational database access through the OPeNDAP Hyrax data server. Exposes databases accessed through ODBC (PostgreSQL, MySQL, SQLite, MongoDB) as standard OPeNDAP datasets.
This project bridged the gap between enterprise databases and scientific data discovery workflows, enabling transparent access to geospatial and climate data stored in relational databases.
FAO's STAC-native earth observation discovery and access platform. Built on STAC-FastAPI with Elasticsearch / OpenSearch, exposing OGC STAC APIs across agricultural indicators, dekadal time series, and administrative boundary hierarchies.
I introduced STAC FastAPI Elasticsearch/OpenSearch (SFEOS) at FAO and contributed patches and design suggestions upstream. This work connects STAC discovery with FAO's Agro-Informatics Platform and catalogue services.
The portal is the Organization's entry point for EO-aligned data discovery — complementing DynaStore's generic geospatial catalog with STAC-native search, item retrieval, and OpenSearch-backed faceting.
FAO's unified identity and access control platform. My 2023–2024 implementation combined Keycloak and Open Policy Agent (OPA) with GCP IAM and integrated Keycloak through a Workforce Identity Pool, supporting federated access across application and resource boundaries.
My contribution covered identity integration, access-policy architecture and implementation in collaboration with FAO teams.
Suite of 5 open-source CKAN extensions for open data platform enhancement, metadata management, and visualization integration.
I provide technical leadership for FAO’s ISO metadata catalogue, implemented with CKAN for the Agro-Informatics Platform, and participated in the standardization work on ISO/TR 19115-4. This Technical Report is listed as under publication in October 2026.
These extensions extend CKAN's capabilities for geospatial data publishing, metadata standardization, and modern web-based data discovery interfaces.
I developed ckanext-jsonschema to support JSON Schema-based metadata forms and validation in CKAN, with documented Solr indexing for schema-related metadata and views. Separately, I worked on an event-driven indexing design spanning Elasticsearch, Solr, and object storage, with GCP Eventarc in the event path. That work is outside the public extension repository.
Tellurion is independently created and owned by Carlo Cancellieri: a self-hosted Rust geospatial serving engine with published v0.6.1 Community source. It covers Features, Tiles, Maps, Styles, STAC, CQL2, COG, Zarr and 3D Tiles; GeoPackage stays the embedded default and PostGIS supports larger deployments.
From a public URL to a map, in a 90-second walkthrough. Explore ESA WorldCover land cover (COG), Monaco building footprints (GeoParquet) and Natural Earth coastline (ZIP Shapefile). See source inspection and map preview with real public data, then try a supported file of your own. This edited walkthrough is not a performance benchmark; previews are temporary and do not create persistent tenants or catalogs.
One bounded ESA WorldCover Item, preserved with provenance and served through split GeoPackage and COG lanes as inspectable STAC, Features, TileSet and live PNG resources.
Public evidence with explicit boundaries:A Python MCP context service that brings relevant skills and task evidence into coding sessions. It keeps the original request intact, scopes retrieval to the project, and shows where each result came from.
o200k_base tokens for serialized MCP schemas and instructions, down from 16,362 for the 87-tool profile (92.41% lower). This is a schema comparison, not a measure of total task usageInteractive retrieval needs no local model or API key. Optional indexing and memory curation run outside the prompt path.
The curated project collection documents Planner, an iOS activity-planning app with LLM-assisted discovery, map-based planning, and local storage.
The curated project collection also documents a Raspberry Pi media-server build, a compact systems and operations project.
At GeoSolutions, I contributed to JMS-based GeoServer clustering and JSON representations of GeoServer REST APIs, working across core code and plugins. This combined distributed-systems engineering with open-source interoperability and API development.
My GeoSolutions work (2010–2014) included geospatial software and data services for institutional and research users. In 2012, I contributed to EOMAP infrastructure supporting GMES Masters and ESA App Camp, making Earth-observation raster time series accessible through OGC services, and to the LaMMA geoportal for meteorological and satellite data.
I also contributed to geospatial infrastructure for NATO NURC/CMRE in La Spezia. The publicly documented company collaboration covered oceanographic data discovery, ingestion and controlled access. This was software engineering for a research and security stakeholder, not military service or a claim of security clearance.
At SIGEO in 2018–2019, I led R&D and engineering delivery for TAP/SMAI, an environmental monitoring system supporting the Trans Adriatic Pipeline construction site in Italy. This connects my platform and delivery experience to civilian infrastructure monitoring.
Public CMRE/GeoSolutions project context · Career and project history
FAO projects are institutional work delivered with colleagues and collaborators. This is a personal portfolio; views and independent projects do not imply endorsement by FAO or any other organization.
My public repositories cover geospatial services, CKAN extensions, catalog integrations, developer tools, and independent experiments. The catalog also contains forks and archived upstream projects; individual project descriptions identify my contribution where documented.
Food and Agriculture Organization of the United Nations (FAO)
Feb 2020–Present | Rome, Italy (Remote)
Architect and lead engineer for the Organization's cloud-native geospatial platforms in support of the Hand-in-Hand Initiative. Technical lead for the Data Catalogue and its integration services, including STAC APIs, DynaStore vector ingestion and serving, and backend services for the Data Manager and Remote Sensing Portal. Led architecture and delivery across DynaStore, the Remote Sensing Portal, and Keystone (unified IAM), coordinating work across the Statistics Division, IT, and regional offices. My Keystone implementation covered 2023–2024; my DynaStore programme work began in 2023. Built and operated review and production environments on Google Cloud, with infrastructure-as-code. Led a multicultural team and mentored engineers across staff and consultant categories. Work includes proposals for dimensions and assets interfaces in geospatial data-cube workflows. Contributed to geographic metadata discussions involving ISO 19115-4 and ISO 19115-5. Integrating Vertex AI and LLM-based tooling (Model Context Protocol) into the Organization's geospatial data workflows.
SIGEO S.r.l.
2018–2019 | Rome, Italy
Led research and development for TAP/SMAI, a real-time environmental monitoring system for the Trans Adriatic Pipeline. Coordinated engineering delivery, Agile practices and scope with consortium partners.
FAO Statistics Division (ESS)
July 2015–December 2018 | Rome, Italy
Senior technical lead on the Statistical Working System, FAO's corporate statistics platform supporting Food Balance Sheets, Supply Utilization Accounts, and domain-specific methodologies across member countries. Led the frontend modernization (React/Redux/TypeScript) and built R-based statistical processing pipelines; established CI/CD and SDLC processes with the SWS team. Role responsibilities also covered testing and validation, QA and production platform support, and technical input to operational handover and maintenance.
Proxima Informatica
May 2014–Jun 2015 | Rome, Italy
Senior Software Engineer from May to November 2014, then Senior System Integrator from November 2014 to June 2015. Client-facing enterprise engineering for major organizations in the international lottery and payments industry. Led Jenkins-based CI/CD and continuous delivery practices within a multicultural engineering team and contributed to the Apache Camel open-source project.
GeoSolutions S.A.S.
Nov 2010–Apr 2014 | Viareggio, Italy
Core contributor to GeoServer, the leading open-source geospatial server, with concurrent Product Owner responsibilities. Delivered solutions for international organizations including FAO, NATO, and NURC. Authored GetLegendGraphic JSON output and contributed to JMX, REST, JMS, and WPS interfaces. Created the GeoServerStressTest load-testing tool and worked on GeoServer Manager REST administration.
OPeNDAP / Google Summer of Code programme
Apr 2009–Aug 2010 | Remote (International)
Selected twice (2009, 2010) for Google's international open-source programme under the OPeNDAP organization. Authored the Hyrax SQL Handler — a C++ BES module enabling SQL/ODBC database access through the OPeNDAP protocol. The public OPeNDAP repository retains the source and development history.
Independent teaching
September 2008–March 2010
Taught Java 5/6 and J2EE 5 courses through classroom and laboratory instruction, preparing teaching materials and practical exercises.
CIRA
January–July 2008
Worked on DataStorageSystem for geospatial and scientific data, using Java 6, XSD, C++ and OPeNDAP.
Ra.Gi.Vi. sas
December 2005–March 2006
Developed ProGM with Java and PostgreSQL, supporting migration and operator training.
A platform includes its deployment, diagnostics and recovery paths. FAO cloud operations and SWS delivery practices connect architecture to the work of running it.
The OPeNDAP database bridge and GeoServer work start from an interface people can use. Standards are useful when implementations make data accessible across tools.
Native code, grid computing, Python services and cloud infrastructure solve different problems. I choose around resource limits, integration needs and maintainability.
Reviewable decisions and shared understanding support delivery. Coordinating, reviewing and mentoring are part of the engineering work.
Architecture, geospatial systems and engineering decisions.
Explore the knowledge notebook: connected notes on architecture decisions, platform operations and geospatial evidence.
Latest articles from WordPress.
16 Sept 2026 · The scarcest resource in an agentic coding session isn’t compute, and it isn’t API budget. It’s context window. Every tool description, every skill definition, every intermediate result an agent has to reason about competes for the same finite space the model
2 Sept 2026 · A dataset that changes its URL every time it moves between storage tiers, gets reprocessed, or migrates catalogs isn’t really a dataset — it’s a moving target that happens to have data attached. GeoID is the persistent identifier service I designed to fix that
19 Aug 2026 · As of this writing, the staging environment for the platform I lead the engineering of at FAO conforms to 77 OGC conformance classes spanning 12 standard families: STAC API (14 classes), OGC API Features (12), Environmental Data Retrieval (5), Processes (5), R
5 Aug 2026 · Multi-tenancy in a database is usually framed as a choice between two bad options: give every tenant their own database (operationally expensive, hard to scale past a few hundred tenants) or put everyone in the same tables behind row-level security (cheap, but
22 Jul 2026 · A few weeks ago I wrote about what makes a GeoTIFF “cloud-optimized” — internal tiling, overviews, a byte layout designed for HTTP range requests instead of sequential reads. That’s the format story. This is the deployment story: what it actually looks like to
17 Jul 2026 · Only 37% of employers still treat a degree as a reliable indicator of talent. That number gets quoted as progress, and it is. But it hides what replaced the degree. Nothing evaluates skill directly. Something evaluates a compressed representation of skill, and
15 Jul 2026 · Two people can sit ten feet apart, each with a capable AI assistant open in a side panel, and learn nothing from each other all week. That is the quiet failure mode nobody budgeted for. We worried AI would replace engineers. The more immediate problem is that
2026 · A practical case study in independent COG pixel validation, explicit provenance and measured range delivery.
2012 · A deployment note from work at the intersection of Java application servers and geospatial infrastructure.
2026 · Shared catalog foundations and the interfaces that project them into different protocols.
2026 · Iceberg, DuckDB and Elasticsearch, with capability contracts that make their differences explicit.
2026 · Engineering judgment and accountability when AI assists with implementation.
2026 · How teams develop technical understanding alongside assisted delivery.
2026 · An AI research experiment, its quality gate and the limits of a promising result.
Open-source work connects this career: OPeNDAP, GeoServer, CKAN and Apache Camel, alongside geospatial standards discussions. Contribution and collaboration are described in the project and career entries above.
Explore public repositories and contribution historyNeapolis University Pafos (Cyprus)
July 2024 | First Class Honours
Master of Business Administration. Thesis on ethical AI integration and ESG frameworks for sustainable business transformation. Coursework in strategic management, finance, operations, and organizational behaviour. Separately awarded Academic Distinction for an Outstanding Academic Record. Read the responsible AI and ESG research case.
University of Naples "Parthenope"
Graduated 18 February 2009
Laurea in Informatica (Class 26). Thesis: "Management of Distributed Data in a Distributed Computational Environment (Grids)" — designed and implemented the data access layer for OPeNDAP Hyrax, leading to Google Summer of Code selection in 2009 and 2010.
Cloud, software delivery and quality
Selected course completions complement my engineering work. The Alibaba Cloud associate certification is historical; its latest validity ended 27 April 2025.
Professional communication and language qualification
Italian: native. English: C1 (self-assessed), based on professional work in English at FAO and an English-medium MBA. Spanish: DELE B1, Instituto Cervantes; examination November 2024, diploma February 2025.
Interested in software engineering, platform architecture and technical leadership? Connect with me about systems, teams and engineering work that needs to last.