Freelancer profile translated to English.

Description

Data & AI Engineer with 4 years of experience in designing, industrializing, and maintaining data pipelines and AI solutions in a production environment.

I have supported major groups like CMA-CGM, PSA, and CAGIP in modernizing their data platforms:

Large-scale data ingestion, transformation, orchestration, and exposure.
On the AI side, I work on MLOps, LLM security, and AI agent design. From the R&D phase to deployment.

What I bring to your projects:

Reliable, tested, and maintainable pipelines. Industrialized AI solutions ready for production.

A capacity to communicate with both business and technical teams.

Technical Stack:Python · PySpark · SQL · Dataiku · Snowflake · Airflow · dbt · AWS · Docker · Terraform · GitLab CI/CD · Power BI · LLM · AI Agents · RAG

Languages

French
Native or bilingual
English
Fluent
Arabic
Native or bilingual

Workplace preferences

Can work on-site

Paris (up to 50km), Nice (up to 50km), Marseille (up to 50km)

SYNCHRONE
Design and deployment of a multi-agent conversational AI platform
DIGITAL AND IT
March 2026 - Today (3 months)
Paris, France
I participated in the design, development, and deployment of Synapse, an enterprise conversational AI platform based on a microservices architecture composed of 7 containerized services, orchestrated via Docker Compose and exposed behind an Nginx reverse proxy at a single entry point.

The core of the platform is an intelligent conversational agent built with LangGraph and GPT-4o-mini, equipped with persistent per-session memory stored in PostgreSQL, capable of maintaining a coherent conversational context over time.
I integrated a RAG (Retrieval-Augmented Generation) system allowing the agent to query an internal document base via vector embeddings, relying on pgvector, sentence-transformers, and ChromaDB for indexing and semantic search.
The platform also includes a module for automatic transcription of audio and video files (WhisperX, Faster-Whisper) coupled with a diarization system (Pyannote.audio) to automatically identify different speakers in a recording.
To ensure security, I implemented a dedicated authentication service based on JWT with refresh token rotation and fraudulent reuse detection.
Large file processing is delegated to asynchronous Celery and Redis workers, ensuring a smooth user experience even for long processes.
The React frontend is served via Nginx, which also handles routing to all backend microservices.

Skills:Python · FastAPI · LangChain · LangGraph · React · PostgreSQL · pgvector · ChromaDB · Redis · Celery · Docker · Docker Compose · nginx · OpenAI API · HuggingFace · WhisperX · Faster-Whisper · Pyannote.audio · sentence-transformers · JWT · Microservices architecture · RAG · Conversational AI
RAG Generative AI AI Chatbot AI Agents
Crédit agricole CAGIP
Automation of ESG pipelines and environmental reporting on Dataiku & Power BI
BANKING AND INSURANCE
August 2025 - March 2026 (7 months)
Paris, France
At CAGIP (Crédit Agricole Group Infrastructure Platform), I worked as a Data & BI Engineer on an environmental and ESG compliance reporting project.
I designed and automated data pipelines on Dataiku for calculating the carbon footprint of the group's internal applications. These pipelines covered the ingestion, transformation, and validation of production data, significantly reducing manual processing and improving the quality of output data.
Orchestration and monitoring of data flows were handled via Airflow and Dataiku scenarios, ensuring traceability and reliability of all processing.
I implemented systematic quality controls and CI/CD processes to industrialize deployments.
I modeled and designed Power BI dashboards for visualizing ESG indicators, providing environmental reporting teams with clear, actionable metrics compliant with group requirements. Special attention was paid to the UX/UI of the dashboards to facilitate adoption by business users.
I also contributed to data compliance with Crédit Agricole group's environmental reporting requirements.
Skills:Dataiku · Python · PySpark · SQL · Power BI · Docker · CI/CD · ESG Reporting · Carbon Footprint · Data Quality · Data Governance · Data Modeling · Data Visualization
Dataiku ESG Data Visualization Spark Python
Capgemini Invent
Design and deployment of AI agents for conversational system security assessment
DIGITAL AND IT
January 2025 - July 2025 (6 months)
Paris, France
As part of an AI security R&D project at Capgemini Invent, I designed and developed, within a team of 4, intelligent agents capable of simulating prompt injection attacks on enterprise conversational systems.

The objective was to assess their robustness, identify vulnerabilities, and qualify associated risks to propose corrective measures.

The agents were developed in Python using the Google ADK (Agent Development Kit) framework, allowing the modeling of different attack scenarios and measuring the resistance of target models against hijacking attempts.
I set up the complete MLOps infrastructure for the project: service containerization via Docker, deployment on AWS ECS, and implementation of a CI/CD pipeline via GitHub Actions to ensure reliable and reproducible deliveries. Code quality was ensured by systematic unit tests with Pytest and rigorous version control.
For operational monitoring, I developed a real-time supervision interface with Streamlit, allowing monitoring of agent behavior, visualization of the detected attack rate, and mapping of identified vulnerabilities. Results were stored in MongoDB.
This project resulted in a complete mapping of risks related to prompt injection attacks and a set of concrete recommendations to strengthen the security of enterprise conversational systems.
Skills:Python · Google ADK · Streamlit · MongoDB · Pytest · GitHub Actions · Docker · AWS ECS · CI/CD · AI Security · Prompt Injection · MLOps · Intelligent Agents · Robustness Assessment
Cloud AWS AI Agents Python Docker AI Security