Cross-platform mobile app

MileageSync

A hands-free mileage logger that detects drives in the background and turns them into IRS-ready deductions, with every GPS point kept on the device.

Overview

A mileage deduction needs a contemporaneous log, and manual logging fails the moment someone is busy — the drives that go unrecorded are exactly the ones worth money.

Automating it means running location tracking in the background, where modern iOS and Android actively work against you. Processes get killed, battery optimisation throttles updates, and a tracker that quietly stops is worse than no tracker at all, because the user only finds out at tax time.

MileageSync puts the hard part behind pure functions. A detection engine reasons over location samples to decide when a drive genuinely starts and ends, and because it is pure it can be tested against simulated restarts, GPS drift, and traffic. Around it sits a background pipeline that persists its own state, self-heals when the OS interferes, and reports its health honestly.

The design principle running through it is that the app should never present a number the user could not defend to the IRS. When trips are unclassified, the deduction says so.

Status: In development — store listings in progress
Built for: Self-employed drivers, contractors, and anyone claiming a mileage deduction or reimbursement
Role: Solo full-stack mobile developer: detection engine, background pipeline, tax and export engine, store compliance
Timeline: Sep 2025 – Present
Platforms: iOS, Android
Categories: Mobile Apps, Location & Maps

Original portfolio entry (opens in a new tab)

The MileageSync app shown on a phone, displaying its trip tracking home screen

What was built

Responsibilities and core features

Responsibilities

Trip detection engine written as pure, testable functions
Background location pipeline that survives process death and OS interference
Tax engine producing IRS-style logs and filing-ready PDF summaries
On-device SQLite schema and migrations
App Store and Play Store compliance, including tracking transparency and purchases

Core features

Background GPS trip detection using speed, distance, and idle heuristics
Manual trip entry for drives the phone missed
A tracking-health card that flags permission and battery problems before they cost a trip
Route previews with a capture-confidence score on every drive
One-tap classification, Needs Review sorting, and bulk classify by week, month, or year
Date-range and trip-type filters driving CSV and PDF exports
Business-use percentage and IRS standard-mileage deduction, marked provisional when trips are unclassified
Per-client and per-vehicle breakdowns
A diagnostics console with live trip state, filterable event logs, and background-task re-registration
One-time in-app purchase to remove ads, with restore for reinstalls

Screens

Trips land without a tap. The health card flags problems before they cost a drive.
Each drive arrives with a route preview and a capture-confidence score.
Older drives collapse into week, month, and year groups for bulk classification.
The deduction is marked provisional while trips remain unclassified.
Filters drive CSV and PDF exports for compliance and finance.
The diagnostics console is the first stop when a QA report comes in.

Engineering

Problems that came up, and what fixed them

The specifics matter more than the feature list. This is what the work actually consisted of.

Making the untestable part the best-tested part
Deciding when a drive starts and stops is the whole product, and it is nearly impossible to test against real hardware. The engine is written as pure functions over location samples, which puts roughly 100 Vitest cases around behaviour that would otherwise only surface in a car — simulated killed-app runtime restarts, GPS-drift rejection so a parked phone does not invent a trip, and red-light tolerance so stop-and-go traffic does not split one drive into five.
Surviving the operating system
Background execution on modern mobile is a negotiation, not an API call. Engine state is persisted so it survives headless JS runtimes and process death, Android 12+ foreground-service restrictions are handled with deferred profile switching, dead SQLite handles are detected and recovered, and an AppState watchdog repairs the background task without the user ever knowing it broke.
A tracker the user can actually trust
A tracker that silently stops is worse than no tracker, because the user only finds out at tax time. A live tracking-health score combines permissions, background task registration, battery optimization state, and GPS heartbeat into one signal with one-tap fixes, backed by a diagnostics console with filterable event logs so a problem can be diagnosed from the device rather than guessed at.
A tax engine, not a mileage total
Raw miles are not what gets filed. The engine computes per-year business-use percentage, applies the IRS standard-mileage rate, and breaks results down per client and per vehicle, then emits an IRS-style CSV mileage log and a branded PDF summary. When trips remain unclassified it marks the deduction provisional instead of presenting a figure the user cannot defend.
Privacy as architecture rather than as a policy
Location history is the most sensitive data the app touches, so it never leaves the device. Trips and GPS points live in on-device SQLite, and data moves outward only through user-initiated share actions on an export the user asked for.
Store compliance designed for, not retrofitted
App Tracking Transparency is prompted only for ad-supported users, SKAdNetwork identifiers are declared, background-mode declarations match the shipped behaviour, and a restore-purchases flow handles reinstalls.

Technical architecture

Expo SDK 54 and React Native 0.81 on React 19 with TypeScript, React Navigation, and Zustand state
A background pipeline over expo-location, task-manager, background-fetch, and notifications, with an Android foreground service and boot-completed re-registration
On-device SQLite with migrations for trips, waypoints, and diagnostics, persisted so tracking survives process death and reboot
Mapbox route matching for road-aligned geometry, degrading to raw waypoints when no token or too few points are available
A Vitest suite spanning the detection engine, storage migrations, tracking services, route matching, tax reports, and exports
EAS Build with sanitized per-environment native config, Google Mobile Ads, and expo-iap for the one-time ad-removal purchase

Technology stack

React Native
Expo
TypeScript
SQLite
Mapbox
Zustand
Vitest
EAS Build

What it adds up to

Roughly 100 automated tests covering detection, storage migrations, tracking services, route matching, tax reports, and exports
Location history never leaves the device — exports are user-initiated only
Produces both an IRS-style CSV mileage log and a filing-ready PDF summary
Background tracking self-heals after process death, reboot, and OS throttling

Related services

The work this project demonstrates

Custom Software

One system that matches your real process, instead of four tools and a spreadsheet holding the gaps together.

Mobile Apps

Cross-platform apps in React Native and Expo — including the background, permission, and store-review work that decides whether an app actually ships.