Portfolio

Context

Design, code, secure and ship Lirix solo, from idea to stores: cross-platform app (iOS, Android, Web) from a single TypeScript codebase, Fastify + Prisma backend, BullMQ cron workers, full CI/CD and automated publishing to the App Store and Google Play. Product positioning: no infinite feed, no ads, no tracking.

Results

• End-to-end delivery in ~8 weeks: mobile (Expo/React Native New Arch), web (SPA), API and workers in production
• Gamification 2.0 — XP, levels, badges, claim-based quests, 30-day reading duels between friends
• GDPR by design: instant soft-delete + automatic purge after 30 days, SHA-256 hashed phone numbers for contact matching, OAuth PKCE, JWT + JWKS, strict CSP
• Automated distribution: every push → lint, tests, DB migration, API/workers/web deploy + mobile OTA via EAS Update

Technologies

Context

Building a SaaS platform for veterinary clinics: two distinct applications (practitioners and pet owners), a native conversational AI concierge on messaging, and a public API for third-party integrations. Differentiating positioning: the pet owner as a first-class user and AI as a native component — the platform coexists with existing PMS systems instead of replacing them. Animal health data and personal data → security by design and GDPR compliance from day one.

Results

• Two-sided platform delivered: practitioners app (records, appointments, note validation) and owners app (tracking, clinic conversations)
• Native conversational AI concierge on messaging — knowledge base with cache pipeline, exposed via an MCP server with a deliberately restricted tool surface
• REST API v1 for third-party integrations: OAuth2 client credentials, per-clinic granular scopes, HMAC-signed outbound webhooks
• Security by design: Supabase with RLS and dedicated Postgres roles, least-privilege throughout, multi-domain deployment on OVH

Technologies

Context

A group of two complementary construction SMBs — a thermal engineering firm (energy audits) and a project management firm — with a shared administrative assistant and four co-CEO decision-makers. Goal: identify AI use cases with real impact, set a data sovereignty framework before any deployment, industrialize in batches validated ROI by ROI. Pragmatic approach: capitalize on the existing stack (Google Workspace, ERP, domain tools) instead of stacking subscriptions, with scopes explicitly ruled out (chatbots, image generation).

Results

• Field audit: one-on-one interviews covering every role — thermal engineers, project managers, admin, leadership
• Recommendations booklet delivered: IT and process mapping, GDPR framework and sensitive-data routing (third-party / sovereign / local), prioritized use cases with quantified ROI and explicit assumptions, technical architectures, phased roadmap
• Three batches deployed: voice pipeline for field notes → transcription → AI → Word reports, automated document control (vendor quotes vs project specs CCTP/DPGF), inter-tool automation from client portal → ERP → shared calendar
• Sovereignty framework set before any deployment: French sovereign AI (Mistral) and local open-source tools (Whisper) for sensitive documents, minimal cognitive load for the teams

Technologies