geomermaids · Cloud-native geospatial experts
Open Source by default, Cloud-native by design.
Architectures, Software, Pipelines and AI applied to geospatial.
Who I am
I design and build cloud-native geospatial systems: pipelines, APIs, dashboards, and ML workflows, using open source wherever it fits. The goal is always the same: geospatial data that is cheap to query, easy to reuse, and independent of any single vendor; geospatial systems that are actually useful, not just technically impressive.
I started out in France contributing to national and regional Spatial Data Infrastructure (SDI) projects, and have spent the last decade on cloud-native geospatial formats and the tooling around them. I now run Geomermaids LLC from the Boston area, working with teams that need geospatial data to hold up under real load and real queries.
My main stack is Python and SQL, with extensive PostgreSQL/PostGIS experience and a more recent focus on DuckDB, to which I've contributed patches. I'm equally comfortable extracting insight from Cloud Optimized GeoTIFFs and multispectral imagery when a project calls for it. I default to object storage (S3, R2) for the data layer and adapt compute and orchestration to fit the problem, with a soft spot for open source geospatial formats and the ecosystems around them.
One thread runs through all of it: these formats and tools work best when they reach beyond the early adopters, and only make sense if they are easy to pick up at scale. That's the gap I'm trying to fill, by building in the open, so others can learn from the work and take it further.
Services
Through Geomermaids LLC, I take on consulting engagements, from an architecture review to building and handing over a working system.
Geospatial data management
Pipelines that turn raw sources (OSM, shapefiles, databases, portals) into clean, typed, partitioned datasets, and keep them fresh on a schedule.
Data publication & visualization
Publishing data people can actually reach: GeoParquet and COG on object storage, STAC catalogs, APIs, and maps or dashboards on top.
GeoCloud architecture
Object storage, serverless compute and CDN caching sized to the workload, so the stack costs little at rest and scales when queried.
Machine learning for geospatial
Feature extraction, classification and change detection on imagery and vector data, with training and inference that run over cloud-native data.
Earth observation & imagery
Sentinel, Landsat and aerial imagery turned into usable products: Cloud Optimized GeoTIFFs, STAC catalogs and on-demand tile services.
Open source migration
Moving off proprietary GIS without losing capability: PostGIS, QGIS and cloud-native formats in place of closed stacks and per-seat licences.
Featured projects
What I'm building in the open right now. See all projects.
GeoPQ Workbench
A free desktop app to open, inspect, fix and query GeoParquet, local or straight from a URL. Grades every file against best practices and rewrites slow ones in one click. macOS, Windows, Linux.
Great datasets, in GeoParquet
Free, queryable straight from a URL: daily OpenStreetMap for North America, OSM infrastructure (power, telecoms, oil and gas, water; preview), FAO GAUL 2024, geoBoundaries and Corine Land Cover.
Get in touch
Have a geospatial project, dataset, or architecture question? Drop me a note.