Full-Stack Engineer (Backend focus)

Role Overview

If you want to do serious engineering on hard Earth-science problems, this is the job. We're looking for a full-stack engineer, strongest on the backend, to take our data platform from working prototype to production, and keep it running.

What you'll work on

Because we pull together so many kinds of science, the data is scattered and sparse. Some of it is scanned maps from decades ago, and it arrives in about a dozen coordinate systems. Your job is to make it fast and get it right, so people can trust what they're looking at. In practice, you'll be:

• Building the ingest workers that take files in, fix or flag their coordinate systems, convert them to GeoParquet, Cloud-Optimised GeoTIFF or Zarr, and record where every version came from
• Keeping the store honest: immutable versions on S3, a catalogue that points at the current one, a review step before anything goes live, and rollback that's just a pointer move
• Getting data in front of people through the API, pre-built map tiles behind a CDN, and 3D views of subsurface models
• Writing the analysis workers that compute statistics from the original data, not a resampled copy, using DuckDB, xarray and rasterio
• Making sure licence and sensitivity rules hold everywhere data leaves the system
• And the unglamorous work that keeps all of us sane: CI on every push, infrastructure as code, monitoring, backups we've actually tried restoring, and secret scanning.

What We're Looking For

You've built backend systems, run them in production, and been the one who got called when they broke. Beyond that, we'd like to see most of these:

• Four or more years of backend work that real users depended on
• Python you're very comfortable in, including async services and files too big to fit in memory
• MongoDB or another document store, plus solid SQL, and a feel for when data belongs in files rather than a database
• AWS in practice: S3, queues, containers and a CDN, ideally set up as infrastructure as code
• Enough JavaScript or TypeScript to ship a frontend feature on your own, maps included
• Good habits: you write tests, you care about CI and code review, and you write down why you decided something
• A liking for vague problems. You'd rather measure than guess, and you say plainly what you found

Nice to have

None of this is required. If you've done some of it, you'll just get going faster.

• GDAL, PROJ, rasterio, GeoPandas or Shapely, and knowing why datums matter (vertical ones too)
• GeoParquet, Cloud-Optimised GeoTIFF, Zarr or NetCDF, and STAC
• DuckDB's spatial extension, xarray or H3
• Vector tiles, deck.gl or MapLibre, and 3D in the browser
• Time spent with geophysical grids, geochemistry, borehole logs, or satellite and hyperspectral imagery
• Work where lineage and reproducibility really mattered
• PostGIS

How we work

• Building with Velocity. We move fast but for us that means less friction, not rushing. You'll own a lot and ship early: your first change should reach production within a few weeks, and you'll own whole services soon after.
• Kalaakari & Kaarigari. We care about craft. A map that never quietly drops a small feature, a pipeline that explains why it rejected a file, an error a geologist can act on. 'It works' is where we start, not where we stop.
• The Science Behind Everything. We write decisions down with the numbers behind them, and we don't call something fast until we've measured it. The people using what you build are geologists, geochemists and ML researchers, and they'll tell you quickly when something doesn't add up. Our numbers end up in front of auditors and partners, so we'd rather a page be slow than quietly wrong.

Disclaimer: This job posting has been aggregated from external source. Role details, content, and availability are subject to change. Applicants are advised to confirm the latest information directly on the company website before applying.