Case Study - Consumer travel app: Wrangle turns camera-roll photos into guides

The problem

Everyone comes home from a trip with hundreds of photos and nothing to show for them. The good shots are buried between screenshots and duplicates, and writing up "where we ate, what we saw" is a chore nobody finishes. Wrangle's bet is that the camera roll already contains the guide. The app's job is to find it.

That is a hard product to build well. It needs a fast, native-feeling mobile app, a backend that can handle large photo uploads, and an AI pipeline that makes good choices about which photos matter and how to group them, without the user doing any tagging.

What we shipped

A cross-platform mobile app built with Expo and React Native, with development and release builds produced through EAS.

An API in NestJS on Fastify, with shared Zod contracts so the app and the API agree on every request and response shape.

A Python worker for the image and AI jobs. Its newest feature, AI Auto Discover, looks through a trip's photos, picks the strongest ones, plans the guides, and drafts each one with a tagline and per-photo captions for the user to review.

Supabase for auth, storage and Postgres, so the team spends its time on the product instead of on infrastructure.

How we built it

Bhimesh designed the architecture and owns most of the codebase: a pnpm and Turborepo monorepo with the mobile app, the API, the worker and the shared contracts side by side. One repository and one set of types means a change to an API shape shows up as a type error in the app before it ships.

The build runs on the same model ESARC sells. AI agents draft the code, tests and migrations in parallel, and a partner reviews every change before it merges. That is how two people cover a mobile app, a backend and an AI pipeline at once.