Corentin Boutillier

**Context**: PharmaGDD is an e-pharmacy e-commerce site with approximately 18,000 SKUs. Following an infrastructure migration (Google Cloud to OVH), the site lost nearly 35% of its organic traffic, and internal technical teams were unable to identify the cause after three months of investigation. The challenge: understand what Google was actually seeing and recover the lost visibility (and revenue).
**Realization**: Technical SEO diagnosis conducted in read-only mode, cross-referencing three sources of truth: a complete crawl of the site's 17,935 pages (using a Python tool developed for the occasion), Google Search Console, and its crawl statistics. The root cause was isolated, dated to the exact day of the migration switch, and quantified: 40% of Google's crawl budget was wasted by redirects on catalog images (the thumbnail generation endpoint returned a redirect instead of the image), which progressively de-indexed pages. A prioritized diagnostic report and a directly actionable technical correction note were delivered to the internal team.
**Result**: After applying the fix (images served directly, without redirects), parasitic redirects dropped from about 20,000 to 100 per day, triggering the resumption of indexing and traffic.

**Context**: Movin'On is the global ecosystem for sustainable mobility (born in the wake of Michelin), bringing together companies, local authorities, and experts around collaborative working communities (COIs), international summits, and a vast editorial library (reports, studies, news, videos). The challenge: replace an old, aging WordPress site with a multilingual editorial platform capable of making this mass of content truly navigable, and offer each visitor a conversational assistant that answers their questions by drawing on the site's actual content rather than a traditional search engine.
**Realization**: Complete design and development of a showcase website, its back-office, and a conversational AI chatbot (RAG). The assistant understands natural language questions, finds relevant content through vector semantic search, and displays corresponding articles, communities, and events in real-time (streamed response) in an interactive panel, including content related to a specific company when the user mentions it.

**Context**: A network of new real estate agents managed their prospects manually, without a suitable CRM to track the sales pipeline, match leads with available properties, or monitor team performance. The challenge: reduce incoming lead processing time and ensure reliable automatic assignment to agents by geographical area.
**Realization**: Development of a custom CRM covering the entire lead management cycle. The Kanban view sales pipeline includes a 24-hour timer and red alerts to ensure responsiveness. A matching engine automatically cross-references prospect criteria (budget, area, tax scheme) with the new real estate property catalog. Automatic lead assignment operates on a round-robin basis by region. The system synchronizes properties and units from two external APIs (Gridky, ImmoDoric) via Symfony Messenger, and qualifies prospects via bidirectional webhooks with QualifLeads. IMAP synchronization of agents' email inboxes centralizes communications within each lead file. The application also includes an interactive Mapbox map, a duplicate detection feature, and token-based API authentication. The project is developed in TDD with PHPStan level 9 and PHPUnit to ensure code reliability.
**Stack**: Symfony 7 · PHP 8.4 · PostgreSQL · Doctrine ORM · Symfony Messenger · Twig · Stimulus JS · Turbo · Webpack Encore · Mapbox · Docker · FrankenPHP · API REST