carderne / gridfinder

Algorithm for guessing electrical grid location based on night-time lights

Hi, I'm Chris

I’m a software/infra/data engineer and love building and breaking and tinkering. If I’m inside but not at the computer, I’m probably reading. And when outside, I’d ideally be climbing mountains but more frequently running and cycling.

I’ve been lucky enough to work across the tech spectrum: designing systems, building apps and services, exploring data, taming models.

By day I write Python (with Pyright set to hard mode) and TypeScript (tRPC + Next.js aren’t so bad), while by night I enjoy Go and Rust and playing with LLMs (because who doesn’t).

I’ve worked with some hot startups in the finance, climate and energy spaces, supported top research institutions and multinationals.

gridfinder's Issues

Performance (numba)

@sebastiantimwagner
I never really found performance to be an issue running at native NTL resolution (~500m), even for multiple scenarios of huge countries (on beefy machines). But should be pretty straight-forward to wrap the optimise() function in @numba, would need to clean it up a bit first though (global state and things like that).

Some of the functions in prepare.py can be annoyingly slow, and many could probably be trivially sped up by using GDAL/OGR directly. But depending on what kind of scenarios, you're unlikely to run these many times so I never really bothered to look into it.