Final degree project, rebuilt
NILM Edge Platform
Rebuild of a university NILM/AI project as a reproducible edge computing and data infrastructure platform.
Non-Intrusive Load Monitoring (NILM) is the problem of disaggregating whole-home power consumption into per-appliance usage estimates from a single sensor. This was the focus of my final university project.
The rebuild
The original project was a research prototype — functional but not reproducible or deployable. The goal of this rebuild is to take the core ideas and produce something that is:
- Reproducible — Nix-based environment, pinned dependencies, no “works on my machine”
- Deployable to edge — targeting small ARM boards with constrained resources
- Properly instrumented — structured data pipelines, experiment tracking, clear evaluation metrics
- Documented — architecture decisions recorded here as the project evolves
Status
Early stages. Currently designing the data pipeline and evaluation harness before any model work. Updates will appear in the blog and notes as the project progresses.