1
See every feature on a map
Every bench, picnic table, trash can, playground, and sign we've catalogued appears on the township map — colored by condition. Tap any marker for details.
Open the map →The parks of Helena Township, on one page.
A complete inventory of every bench, picnic table, sign, trash can, and playground across Helena Township's parks — each with a photo, a location, and a rated condition. Built for the people who live here, for the committee that stewards these places, and as a working example of how modern edge computing can serve a small township on the shore of Torch Lake.
—
features catalogued
—
need attention
—
parks covered
—
reviewed and confirmed
Three things this inventory does
For residents, for the committee, and as a reference architecture.
See what we have
Every bench, sign, picnic table, trash can, and playground feature across five parks, in one place, with photos and locations. Residents and the committee can answer "what's actually there?" in under a minute.
Track what needs care
Condition ratings highlight what's holding up and what needs attention. Good stewardship of the parks we share starts with knowing their current state — and this inventory is a public record of it.
Show how it's built
A working example of edge computing for public good: Akamai CDN, serverless functions, object storage, and a GPU classifier — all wired together to serve a township of about 1,200 people. See the technical details.
Using the inventory
Four ways to explore the inventory.
2
Find what needs attention
One tap on "Needs attention" surfaces everything rated poor or hazard. Filter by park, type, or condition to zero in on the priorities for the next work cycle.
Show the attention list →3
Check the details
Every feature has a photo, a condition rating, and a short reason for the rating from an automatic photo review. Conditions are reviewed and confirmed by committee members over time, and the catalog reflects the latest state.
Browse the catalog →4
Learn how it's built
A short technical tour of how this inventory runs: Akamai CDN in front, serverless edge functions handling requests, a GPU classifier on a Linode cluster, and real-time observability piping back to ClickHouse. A reference architecture for public good.
See how it works →How this inventory was made
- A volunteer walks each park and photographs every feature, GPS-tagged by the phone.
- Photos upload straight to Linode Object Storage via a Spin-based edge function running on Akamai's global network.
- Each photo is classified by Qwen2.5-VL and CLIP running on a GPU in a Linode Kubernetes cluster — photos never leave our own storage.
- The results render here, at the edge, globally — a catalog, a map, and a story about what our parks need next.
Full technical tour: the architecture page has the topology diagram, the stack list, real latency numbers, and a link to the open-source repo.