GIS · mapping · 3D

From GIS to 3D: your geospatial data,
usable at last

Layers scattered across a shared drive, drawings that will not line up, field surveys that never make it into the system. I structure the data first, then produce what needs to come out of it.

Describe your dataset → The four kinds of work
QGIS PostgreSQL / PostGIS GDAL MapLibre GL JS AutoCAD

The starting point

The data already exists.
Its condition is what holds you back

In most organisations there is no shortage of geospatial data: it has piled up. Deliverables from a succession of contractors, partial exports, coordinate systems that do not match, attributes named differently from one file to the next, and three versions of the same network with no date or author.

Producing a map on that basis gives you a wrong map, delivered fast. The useful work starts with the clean-up: checking, realigning, normalising, deciding which source is the reference. It is less spectacular than a rendering, and it is what decides the rendering's worth.

The question I am asked is almost always about format. The right question is the condition of the content.

A map produced from data that has not been cleaned up is a wrong map, delivered fast.

Services

Four kinds of work

They almost always follow one another in this order. Starting with the third gives you an application that displays data nobody vouches for.

01 · Clean-up
Conversion and normalisation, realignment of coordinate systems, topology checks, deduplication, and a data model that holds up over time. Loading into PostGIS when the volume or the number of contributors justifies it — not on principle.
02 · Production
Maps, automated atlases, regulatory map sheets, spatial analyses. Cross-referencing with public reference data: IGN, BRGM, risk prevention plans, local open data.
03 · Applications
Web mapping applications for viewing and data entry, designed to be hosted by you and taken over by your team. They are built on a library I publish under the MIT licence: no proprietary server, no API key to keep paying for.
04 · Drawings and 3D
Roads and utilities (VRD) and site development drawings, AutoCAD automation, CAD-to-GIS bridges, 3D modelling of terrain and structures from surveys or point clouds. This is the kind of work where the input data is the most heterogeneous, and therefore the one where the clean-up matters most.

Case study

When the register outlives the report

Street lighting · consortium · GIS component

SIDELEC Réunion

What I did

Recovery of the existing inventories, structuring of a light-point register that several contractors could work with, integration of the field surveys, production of the map deliverables for the diagnosis.

What remains

A diagnosis is there to decide on a works programme: the register has to live on after the report is handed in. That is precisely the point where most GIS deliverables die.

The case in detail → Keeping an asset register alive →

Clients

The organisations I work with

Profiles

  • Engineering consultancies: roads and utilities, environment, land development, energy.
  • Network operators: water, sewerage, electricity, lighting, telecoms.
  • Local authorities and joint authorities.
  • Construction and civil engineering firms.

How I work

  • Engagements in French-speaking countries and across Europe.
  • Engagements in English are possible.
  • Consortium bids and subcontracting on tenders.
  • Days spread out over time, with no minimum duration imposed.

Questions

What people ask me about data

In practice, anything GDAL can read: shapefile, GeoPackage, GeoJSON, DWG and DXF, KML, geocoded CSV, exports from business databases, rasters and point clouds. The useful question is almost never the format: it is the condition of the content — coordinate systems that do not match, attributes named differently from one file to the next, three versions of the same network with no date or author.
Not always. A database is justified by volume, or when several people need to write to the same data without treading on each other's toes. Below that, a well-structured GeoPackage does the job, and your team can still handle it without an administrator. I suggest a database when it serves a purpose, not on principle.
No, and that is deliberate: they are designed to be hosted by you and taken over by your team. No proprietary server, no API key to pay for again every year, no dependence on my availability. The mapping library they are built on is published under the MIT licence.
Yes, as a consortium member or as a subcontractor. The diagnosis of Réunion's street lighting network was won that way, with the GIS component as my share. Engagements can be days spread out over time, with no minimum duration imposed.
Yes. Engagements can be carried out in English, both written and spoken. Map deliverables and documentation can be produced in both languages.

On the blog, in French

Further reading

Geospatial formats in 2026: GeoJSON, MVT, WFS — comparison & architectures

12 min

Hosted PostGIS in production: deploying, securing, loading data

12 min

LiDAR in practice: formats, reading, QGIS integration and processing

14 min

Describe your dataset

Formats, volume, and what has to come out at the end: enough to scope the work seriously in a single exchange. Reply within 48 hours.

Describe your data →