Freelancer profile translated to English.

Description

Senior AI Product Engineer & Data Scientist | Expert GenAI (Multi-Agents, Advanced RAG) & MLOps

**Looking to integrate Generative AI into your product but worried about the lack of reliability, latency, or hallucinations of LLMs**?

As a **Senior AI Product Engineer & Data Scientist**, I design and deploy robust, deterministic AI systems aligned with your ROI. I specialize in transforming your Proofs of Concepts (PoCs) into scalable, industrial-grade features.

🚀My Added Value & Differentiation

What sets me apart is my dual background in Deep Tech & Software Engineering. Thanks to my PhD from École Polytechnique and 8 years of R&D experience, I architect AI as a strict software component:

Zero Hallucination: Securing Tool Calling and extraction via Pydantic schemas with self-healing logic (JSON auto-correction).
Cost Control: Optimizing pipelines (Prompt Caching) and production tracking via LangSmith.
Production-Ready Code: Properly typed Python, tested, containerized (Docker), and integrated into your CI/CD pipelines.

🛠️Types of Projects and Regular Deliverables

Multi-Agent Architectures: Asynchronous agent pipelines (Agno, LangGraph) simulating complex business logic.
Advanced RAG: Query Transformation modules and vector database indexing (ChromaDB).
Data Science & ML: Dynamic pricing models, heavy time-series processing, and constraint optimization.
Full-Stack MVP Systems: Modern applications connecting your interfaces (Next.js) to secure relational databases (Supabase/PostgreSQL, RLS) with Full-Text search engines.

Available immediately for scoping, AI architecture, or technical reinforcement missions. Hybrid format (Île-de-France) or Full Remote.

Languages

French
Native or bilingual

Workplace preferences

Can work on-site

Paris (up to 50km)

Projet Indépendant / R&D Tech
AI Architect & GenAI Expert - Multi-Agent System & Advanced RAG
May 2026 - May 2026
End-to-end design and deployment of a complex asynchronous B2B multi-agent AI architecture aimed at automating knowledge extraction and business query analysis. Total focus on reliability, determinism, and production observability. Key achievements:
Multi-Agent Orchestration: Development of an asynchronous pipeline of 7 LLM agents (Agno / DeepSeek) simulating business logic and structured synthesis generation.
Advanced RAG: Implementation of a Query Transformation module connected to a vector database (ChromaDB) with optimized micro-chunk indexing to handle dense or fuzzy queries.
Guardrails & Reliability: Securing Tool Calling via Pydantic schemas and implementing self-healing logic (JSON auto-correction) to eliminate hallucinations.
MLOps & Evaluation: Fine-grained tracking of latency, Prompt Caching via LangSmith, and deployment of an automated evaluation framework (LLM-as-a-Judge).
Generative AI Multi-Agent Systems RAG LangGraph Python
Consultante Indépendante / Entrepreneur
Data Scientist & AI Product Engineer - Data Strategy & Full-Stack MVP
March 2022 - Today (4 years and 5 months)
Technical and strategic support for companies in data valorization, business model optimization, and integration of high-performance software and AI solutions. Key achievements:
Data Analysis & Dynamic Pricing (Real Estate): Exploiting market data and developing dynamic pricing models to optimize ROI and occupancy rates for real estate portfolios.
GenAI Automation (E-commerce): Scoping needs and integrating generative AI solutions (OpenAI, Gemini LLMs) for copywriting optimization through advanced Prompt Engineering.
Full-Stack MVP Focus — "C'mon artisan" Platform: Complete design and development of a geolocation platform for artisans (Next.js 14, Supabase, PostgreSQL). SQL implementation of the Haversine formula for real-time geospatial distance calculation and configuration of a French-optimized Full-Text search engine using GIN indexes.
Data science dynamic pricing Prompt engineering PostgreSQL Next.js
Admo.tv
Senior Data Scientist
October 2016 - February 2022 (5 years and 4 months)
Paris, France
R&D and industrialization of the Drive-to-Web attribution engine for the TV analytics leader. Handling massive data volumes and ensuring team production practices. Key achievements:
Time Series Processing: End-to-end design and industrialization in Python of heavy algorithms (peak detection, Lift analysis) for TV impact analysis.
Operations Research: Development of a ranking and optimization algorithm under budget constraints with a direct impact on client ROI.
Engineering Leadership: Technical mentorship of junior profiles and guarantor of production standards (CI/CD, Docker, code reviews, test coverage).
Time Series Machine learning Python Docker CI/CD