Freelancer profile translated to English.

Description

I architect and develop production AI systems — RAG pipelines, autonomous multi-model agents, custom LLM integrations.

Formerly Lead Developer at CERN (real-time robotics interface, 3 years), I now design two SaaS products in production on the Claude API: an AI content generation tool for infopreneurs and a real estate prospecting platform with multi-model AI scoring.

What I bring to your projects:

You have an existing product and want to integrate AI that truly works in production — not a POC that breaks at the first edge case. I design the architecture, choose the right models for your use case, connect your data, and deliver a system that runs.

Specifically, my interventions focus on:

→ AI Agent System Architecture (multi-agent orchestration, agentic loops, error handling and fallbacks)

→ Production RAG Pipelines (ingestion, chunking, embeddings, retrieval, generation)

→ LLM Integrations (Claude API, GPT, open-source models via Groq/Ollama) into business workflows

→ Performant Full-Stack Applications (Next.js, React, Python, Supabase, Vercel)

→ Infrastructure and Deployment (Docker, CI/CD, reverse proxy, monitoring)

Main Stack: Next.js · React · Python · TypeScript · Supabase · PostgreSQL · Claude API · Groq · Vercel · Docker · MCP (Model Context Protocol)

Industry field of expertise

Languages

French
Native or bilingual
English
Native or bilingual
Spanish
Fluent

Workplace preferences

Remote only

Primarily works remotely

Joube.ai
Co-founder & Lead Developer
August 2025 - Today (1 year)
AI content generation SaaS for French creators and infopreneurs. The platform analyzes the user's complete positioning (customer avatar, content pillars, marketing tensions) and then generates personalized Instagram Reels scripts and visual carousels.
Architecture and development of the entire product:
→ Conversational AI Onboarding: A 6-step pipeline that automatically generates the customer avatar, tensions (objections, frustrations, desires), and 3 content topics per tension per pillar
→ Script Generation in 3 Sequential LLM Calls: Content (Claude Sonnet via OpenRouter), then hook/CTA selection by dual ranking (Claude Opus), then revision by user feedback
→ Hook Intelligence Engine (Separate Node.js Express Service): Daily scraping of reels from several expert creators, audio transcription (MLX Whisper), AI segmentation, deduplication using OpenAI embeddings text-embedding-3-small (pgvector, 0.85 cosine similarity), composite engagement scoring → feeds script generation with proven and contextually aligned hooks
→ Content Tracking: Automatic publication detection by comparing embeddings (audio transcription ↔ script + caption ↔ IG description)
→ Carousel Editor: 5 slide templates generated by Gemini 3.1, Canva-like editor with AI chat to modify slides in natural language, PNG export
→ Pre-generation Validation: Kimi K2.5 (Groq) evaluates user-provided context and asks a targeted question if necessary before launching generation
→ Automated CRM: 4 scenarios triggered by user posting activity
Stack: Next.js · Claude API (Sonnet + Opus via OpenRouter) · OpenAI Embeddings · Supabase (self-hosted, pgvector) · Python · Node.js Express · MLX Whisper · Gemini 3.1 · Groq · Bright Data · S3 · Traefik · Vercel
Product in production with paying users. Continuous iteration based on user interviews.
Next.js Generative AI Supabase RAG LLM
CERN
Software Engineer
PUBLIC SECTOR
March 2022 - April 2025 (3 years and 1 month)
Genève, Switzerland
Improved the robotic graphical interface through software development, optimizing operator efficiency and system ergonomics (GUI developer, robotic optimization).
Developed a multi-process Python application, increasing data transmission efficiency by 60% (Python multiprocessing, data transmission).
Designed a web interface and PostgreSQL database for configuration file management, streamlining robotic operations (web interface PostgreSQL, config management).
Led a team of three engineers on a Unity project, delivering a visualization tool two weeks ahead of schedule (Unity leadership, 3D visualization).
PostgreSQL Unity Python Node.js
Bimedoc
Backend Developer
MEDICAL
September 2021 - January 2022 (4 months)
Lyon, France
Implemented CRUD methods with Python Django, improving API response time by 30% (backend developer Django, CRUD API).
Performed code reviews for a team of 5 developers, enhancing code maintainability (code reviews, maintainability).
Conducted penetration tests, resolving several security vulnerabilities to meet industry standards (penetration testing, app security).
Python (Programming Language) Django CI/CD Agile Methodology