sprintcoach.co.nz
Sprint Coach
Nathan’s coaching business grew by word of mouth. His first website needed to show athletes and parents who they would be training with, and make the next step easy.
Work
Things I’ve made.
Client websites, independent software, and research.
BrontëHF
Brontë brought the design. I built a portfolio they could keep updating themselves, with an editing workflow that preserves their visual direction.
You Inc
I wanted one view of what I own, owe, earn and spend. I built a finance product around a ledger, then removed the SaaS layer so it could belong to the person running it.
Idle GPU → billed at cost
Always on Accounts · ledger · invoices Reachable with your PC off
← Outbound tunnel No open ports
Your machine Ollama · Blender Only while powered on
kWh rate×GPU watts=price No markup
GPUShare
My desktop GPU sits idle most of the day. I built a way for friends to share it for local AI and rendering, with the limits of a home machine visible.
Apple Health→Postgres→Trends
HealthAgent
My nutrition and activity lived in different apps. I brought their Apple Health exports together so I could see the trends in one place.
One 5W host · two doors
PublicCloudflare TunnelPhotos · files · calendar
AdminTailscaleSSH · Postgres · SMB
Raspberry Pi 4 · 8GB 7.26 GB allocated · 9.25% headroom
Restore tested1h 45munder a 2h target
Home Lab
What would happen if I lost the machine running my home services? I built a small server around that question, then tested a full recovery.
PySpark experiment1.88M records
- Class G recall
- 0 → 0.7518
- Weighted class G precision
- 0.0099
- Weighted overall accuracy
- 32%
Recorded random-forest results · class weighting finds more rare fires, with many more false positives.
Wildfire PySpark
A model can look accurate while missing the fires that matter most. I explored that problem across 1.88 million wildfire records.