Ridwan HamidBack to the journey

Case study 03

LightStack Solutions

From concept to shipped software.

I co-founded LightStack, a studio that designs, builds and maintains mobile and full-stack applications. Two of its apps are live on the App Store, and both stand on a shared engineering foundation.

Role
Co-Founder, full-stack and mobile developer
Since
Nov 2025
Shipped
Distant and Glucose Tracker, on the App Store

The studio

LightStack builds production iOS and Android apps for clients and for itself, and keeps them running on support and maintenance retainers. The work below is three parts of the same engineering: two shipped apps, and the foundation they share.

Distant: an AI coach that remembers

Distant is LightStack’s private AI relationship coach, live on the App Store. I designed and built the coaching experience from the ground up: the LLM-backed conversation engine, a persistent multi-session memory layer, and pattern detection that surfaces behavioural themes across a user’s past check-ins.

Onboarding is a question graph whose answers compile into the coach’s long-term memory, so the first conversation already knows the person. I also built the core flows (onboarding, sign-in, coaching sessions, conversation history and reflection), UI and backend.

Consent to AI processing is enforced on the server. Deleting an account clears it from every service that holds the user’s data in an idempotent order, the user’s own record last, so an interrupted deletion can always be retried.

Distant’s welcome screen: “Your private relationship co-pilot”.
Welcome.
An onboarding question: “How would you describe your communication style?”, with five answers.
Onboarding answers become the coach’s memory.
Distant’s home screen, with “Begin your relationship journey” and the messages remaining this period.
Home.
A conversation in which the coach draws on what the user shared during onboarding.
The coach recalls what it learned.

Glucose Tracker: a client’s health app

Glucose Tracker brings glucose, meals, insulin and activity into one timeline, for iOS and Android, and it is live on the App Store.

  • Two-way, incremental sync with Apple HealthKit and Android Health Connect, with offline queues and retries, so the data stays continuous in the foreground and the background.
  • AI meal analysis from a photo or a recipe link, refined in conversation.
  • Glucose charts drawn on the GPU with Skia: the day’s profile, and the response after each meal.
  • 1,649 unit tests and 73 end-to-end flows.
A day’s timeline of glucose readings, meals with insulin doses, notes and exercise.
One timeline for the day.
The insights view: the day’s glucose chart with timeline events, and nutrition per meal.
The day’s glucose, drawn on the GPU.
A meal analysed from a photo: a salad broken into ingredients with carbs, protein, fat and calories.
A meal analysed from a photo.
A run with its glucose impact, before and after, and a chart of glucose and heart rate.
An activity’s glucose impact.

The engineering foundation

Every app needs the same things: sign-in, onboarding and consent, subscriptions and paywalls, notifications and account deletion. A module is promoted into LightStack’s shared foundation once a second app needs it.

Each app sets its modules up in config that is validated at build time, with typed client files. The foundation ships as a private package, with a scaffolder that starts each new app on it, and CI scaffolds a fresh app from the packed release on every change. Its database access rules are pinned by tests.

Account deletion revokes the Apple sign-in token first and deletes the subscription last.

Outcomes

  • Two apps live on the App Store: Distant and Glucose Tracker.
  • Delivery of repeatable app builds went from months to weeks.
  • Seven releases of the foundation in five weeks.