One domain — motorsport telemetry — taken from firmware on the car to a board in the browser.
flowchart LR
ESP["<b>TrackPro_ESP</b><br/>C++ · ESP32 / ESP8266<br/><i>on the car</i>"]
APP["<b>TrackPro</b><br/>Kotlin · Jetpack Compose<br/><i>in the hand</i>"]
API["<b>TrackBoard_Backend</b><br/>.NET 10 · EF Core 10<br/><i>on the server</i>"]
WEB["<b>TrackBoard_Web</b><br/>Angular 21 · signals<br/><i>in the browser</i>"]
ESP -- "Wi-Fi TCP · Bluetooth" --> APP
APP -- "sessions · laps · tracks" --> API
API -- "account restore" --> APP
WEB -- "reads · account writes" --> API
classDef built stroke:#14161a,stroke-width:2px
class ESP,APP,API,WEB built
TrackPro_ESP — the firmware, C++ on ESP32 / ESP8266
Reads a GPS module and pushes position, altitude, satellite count, speed and timestamp over Wi-Fi — TCP for the lowest latency, with a WebSocket variant alongside it. Ships a Python simulator so the phone app can be developed and tested with no rig on the bench.
TrackPro — the app, Kotlin and Jetpack Compose
Lap timing with a live delta against the session best, sector and checkpoint splits, sprint timing, and drag timing for the quarter mile and 0–100 km/h. A track builder records custom circuit geometry. Afterwards: lap-versus-lap comparison, speed heatmaps on the map, and a theoretical best assembled from the best sectors.
Three GPS sources are all first class — ESP32 over Wi-Fi, ESP32 over Bluetooth Classic, and the phone's own receiver — so nothing in the app may assume the sample rate of the good one.
It is also the TrackBoard client: sign-in with token refresh, background sync of leaderboard-eligible sessions, photo upload, and a restore path that rebuilds an account's tracks and sessions on a new phone. The phone is the source of truth for the whole ecosystem.
TrackBoard_Backend — the API, .NET 10 and PostgreSQL
Drivers publish the tracks they build, upload a session's laps, and compare best laps on a public per-track leaderboard. ASP.NET Core on EF Core 10 and PostgreSQL, deployed on Render. JWT auth, OpenAPI with a Scalar UI, and split liveness and readiness probes so a deployment can tell the difference between running and ready.
Detail that matters more than the feature list: UUID v7 keys, Mapperly projections so a leaderboard query stays one round trip, and a pinned transitive dependency closing a high-severity advisory with auditing on across the whole graph.
The championship model underneath is built and tested but parked, because nothing TrackPro records maps onto it yet.
TrackBoard_Web — the board, Angular 21 with no UI library
Public: every published track with its outline drawn from driver-published GPS points, a best-lap classification per track on a shareable link, and a page per driver. Signed in: season record and personal bests with board position, every uploaded session broken down by lap and sector, the garage, private tracks, and an account that can export itself to JSON or delete itself.
Standalone components, signals and rxResource, and no UI library at all — the interface is
written, not assembled. Sessions, laps, tracks and cars are read-only here; the web writes
nothing but the account.