CarHeat

Turns "it's hot outside" into a car-specific cabin temperature, before it's dangerous. Overview

CarHeat is a Flutter app (iOS + Android) that predicts how hot a parked car's cabin will actually get, using live weather for the exact parking location combined with the vehicle's own physical properties like paint color, body type, interior material, and window tint. It returns one number with a severity band (SAFE / WARM / HOT / DANGEROUS / EXTREME) and a plain-language warning like "dangerous for kids and pets." The Problem

Standard weather apps report ambient temperature only they don't account for how a black SUV with leather seats in full sun heats up completely differently from an untinted white hatchback in shade, even on the same day. Parked cars can hit lethal temperatures far faster than people expect, and there's no trustworthy signal telling drivers when or how to act. CarHeat closes that gap by turning "it's hot out" into a concrete, car-specific cabin number backed by the physics behind it. Key Features

Parked-Car Heat Estimate (Unique) — Steady-state thermal model combining solar load, body/window heat absorption, interior material, ambient boost, wind cooling, and UV, calibrated against two published heatstroke studies (McLaren 2005, Vanos 2018).

Live Weather via Apple WeatherKit — Real-time conditions for the car's exact location, pulled server-side via a signed JWT.

Apple Watch Companion & Complication (Unique) — Check cabin heat, mark "I parked here," and manage reminders from the wrist; built as a fail-safe native bridge that's inert on non-iOS devices. Tech Stack

Mobile: Flutter · Riverpod (state) · Dart, Swift, Kotlin. Backend: Firebase (Cloud Firestore, Cloud Functions on Node 22/TypeScript, Auth, Cloud Messaging, Remote Config, Crashlytics, Analytics) · REST API. Infra: GitHub · Server-side cron sweep (3-min interval) for reminder dispatch instead of per-job scheduling, to keep costs near zero at scale. Results & Key Takeaways

Removed account creation entirely (was stalling ~2 min on some Android devices) device-based identity restores the same user silently after reinstall.

Every ad placement is remote-configured, not baked into the build, so ad load can be tuned or killed with no app release.

Batched reminder delivery keeps server cost near zero, trading a few minutes of delivery precision for near-zero unit economics.

The core differentiator a vehicle-specific, physics-calibrated cabin temperature rather than a generic weather reading is the clearest thing to lead with in store listing and marketing copy.