About the project

Why I built this

I'm Dan — a CS graduate about to spend 2–3 years travelling across four continents on working holiday visas. Before I left, I wanted to answer a question: what does long-term nomadic life actually look like in data?

TravelNet is the answer. A Raspberry Pi 4B runs unattended throughout the trip, ingesting GPS traces, health metrics, financial transactions, and weather data around the clock. By the time I return, I'll have a continuous, multi-year personal dataset unlike anything that typically ends up in a portfolio.

The ML pipeline — clustering, segmentation, anomaly detection — kicks off from Australia, using the US baseline data for comparison. The public demo here is the anonymised output of all of it.


Architecture

ingest FastAPI on port 8000
dash Flask admin UI on port 8080
proxy nginx, TLS via Tailscale
store SQLite on external HDD
sync Tailscale remote access
bkp Cloudflare R2, age-encrypted

Data sources

GPS Overland (iOS) — 5s intervals
GPS iOS Shortcuts — gap fill
Apple Watch Ultra + HAE
💳 Revolut CSV exports
💳 Wise ZIP exports
🌤 Open-Meteo weather API

View on GitHub ← Home