Selected work

Geowitness

A location-first video app for real events: pick a place and a date on a map, watch what people there actually filmed.

Role
Full-stack developer at TheCodingMachine
Year
2021 to 2022
Status
Live
Stack
FlutterDartSymfonyGraphQLVueMySQLDocker
Geowitness App Store screens: saved locations on a map, event selection, the video grid for one event, the player with comments, and the contribute screen

Inside the app

Geowitness player showing a flooded street in Broward with comments, share and bookmark controls
Watch, comment, share. The player for one clip inside an event.
geowitness.com home page
geowitness.com, the product site

The problem

Social feeds answer "who do you follow", not "what is happening on my street right now". When a road floods or a fire starts, several people film it and the footage scatters across platforms where it is indexed by author and recency instead of by place. The client wanted the opposite: witnessing an event should be as simple as choosing where, when and what, on a map.

What was built

  • A Flutter app for iOS and Android built around a map: search or pin a location, keep a set of favourite places, and scrub a date range to see what was filmed there and when.
  • An event model where the event, not the clip, is the unit. One flood on East Broward Boulevard groups the six videos different people shot of it, sorted by time.
  • A vertical player with comments, share, bookmark and view counts, plus a Contribute flow that records or uploads a clip and attaches the place and date.
  • A Symfony and GraphQL backend over MySQL with a Vue admin panel for moderation, the whole stack containerised with Docker so it ran the same everywhere.

Decisions worth explaining

  • Index by place and time, not by a social graph. A new social app has no follower network to bootstrap from, but it does have a map, so the map does the discovery work that follows would normally do.
  • GraphQL because the map screen, the event list and the player each need a different shape of the same underlying records, and one REST response could not serve all three without over-fetching on mobile.
  • Flutter for a single codebase across both stores, since the interface is map-heavy and near-identical on each platform and the team was small.
  • Docker around PHP, GraphQL and MySQL so the backend was reproducible for a distributed team and for deployment.

Outcome

Live on the App Store since 2022 and still listed, rated 16+ in News, running on iPhone, on Apple Silicon Macs and on Apple Vision. Delivered for the client, who publishes it.

More work