My Work

Most apps fight to keep you on the screen. The thing people actually want from an outdoors app is the opposite: to close it and go do something. But choosing where to go is its own friction. Too many options, none matched to how much time, energy, or motivation you have right now.

A recommendation flow that takes your mood, time, energy, and location and returns nearby walks, parks, trails, viewpoints, and quiet streets, fast. Around it sits the rest of a real product: interactive map exploration, saved favorites, full trip-planning workflows for bigger adventures, and an account-free demo anyone can try. I designed the whole thing first (information architecture, flows, wireframes, high-fidelity screens), then built it end to end. Grasstile supports 11 languages across 15 routed locale variants.

Reached an active invite-only closed beta with an account-free public demo. Planning and account access require an invitation, and purchases are unavailable. The product deliberately avoids streaks, badges, and guilt mechanics. Its goal is to help people choose something worthwhile and then leave the screen.

Building the full stack alone forces ruthless scoping: every feature has to earn the maintenance it costs. Wiring up internationalization (next-intl) and an AI recommendation layer early was far cheaper than retrofitting them, and designing against engagement metrics made the product decisions clearer, not harder.

The brief was a capstone activity tracker, with two real constraints: it had to run as a native mobile app on a tight school timeline, and it had to be fully French, not an English app with translated labels. A tracker also has an awkward demo problem: how do you show live GPS and sensor data to a room full of people who are sitting still?

A React Native app, French-first from the ground up, that reads the device sensors to track live speed, pace, distance, altitude, and elevation gain. Firebase handles auth (email/password, Google sign-in, and an email-verification flow). Around the tracker sits a full app: quick-start session types (walk, run, sport, quick test), daily/weekly/monthly stats, session history, in-app notifications, and a profile area. My answer to the demo problem was a simulation mode that fakes sensor data so the whole flow can be shown indoors, without going for a run. I owned the design and front-end build and split delivery with a school partner.

Delivered as the final projet intégrateur and distributed as an installable Android APK. It works end-to-end: sign in, start a session, watch live sensor data, save it, and review it in history. The simulation mode means anyone can try the full flow from their desk.

My first cross-platform mobile build and my first time wiring up Firebase auth and reading raw device sensors, so much of it was learning the mobile lifecycle, permissions, and managed-auth patterns under a deadline. Splitting delivery with a partner also taught me to scope and hand off cleanly instead of holding the whole thing in my head.