Valentin Boutouria

Context

Each activation (AR mirrors, lenses) produced its own numbers; clients needed one clear interface, readable without a technical background and flexible enough for every project.

Key tasks

• Back end: Supabase / PostgreSQL database with 3 tables (projects, mirrors, visitors) linked by foreign keys, fed by the mirrors on every visit
• Cron jobs that automatically purge collected data
• Vue 3 and TypeScript front end: a modular grid of number (1×2), chart (4×2) and pie (2×2) widgets, added from an extensible catalog and rearranged by drag and drop
• Hand-made SVG charts, usable with the keyboard and readable by screen readers
• Aggregated count queries, with no personal data pulled into the browser, plus 50 Vitest unit tests

Outcome

A complete pipeline, from a visit in front of the mirror to a number on the client’s screen, with a dashboard everyone arranges their own way.

Context

The flow had to work on every smartphone, load fast, match the brand’s visual identity and follow the advertiser’s tagging plan to the letter.

Key tasks

• Built the campaign screens in SvelteKit, mobile-first, with Honda’s fonts and brand guidelines
• Sign-up form validated on client and server, protected by reCAPTCHA, stored in Supabase
• Optimized Three.js 3D scene (Draco-compressed model, HDR lighting) and an animated canvas reveal
• GTM / dataLayer tracking (CTA clicks, form, scroll depth) behind OneTrust consent

Outcome

A single front end chaining landing, creation, video reveal, sign-up and AR game, deployed on Cloudflare.

Context

An organizational goal: give teams a clear view of ongoing incidents (severity, status, owner) so they can prioritize better, anticipate workload and plan their interventions.

Key tasks

• Angular 18 front end in TypeScript: standalone components, reactive forms with validation
• Front-end tests with Jasmine / Karma, back-end tests with JUnit 5 and Mockito, JaCoCo coverage
• Spring Boot REST API (Java 21) with standardized errors, PostgreSQL schema versioned by Flyway
• Multi-stage Docker images, Docker Compose and a GitHub Actions pipeline (build, tests, images)

Outcome

docker compose up --build is all it takes: the front end, the API and a seeded database start together.

Context

Extend the visit into the phone, with no app to install: every exhibited object had to tell its story in just a few gestures.

Key tasks

• “Milano” trunk: scan the trunk, then points of interest on each accessory open their own card (glove stretcher, brush, blotting book…)
• Three generations: scan the portraits of Louis, Georges and Gaston-Louis Vuitton, each revealing a chapter of the Maison’s history
• Art Deco windows: marker tracking and shader-based occlusion, so the 3D slips behind the real scenery
• Shared guided flow: tutorial, contextual hints, tap interactions and localized text

Outcome

Experiences that open with a single scan from Snapchat, built to stay smooth on the widest range of phones.