Freelancer profile translated to English.

Description

Generative AI in production is not just an API call. Expert AI Engineer with 6 years of experience, I focus on the critical transition from POC to massive industrialization. My track record includes deploying RAG (Retrieval-Augmented Generation) solutions for the banking sector, serving 8 million users with a load of 1000 requests/minute.

My approach is tool-agnostic: I don't sell a technology, but a robust, sovereign, and optimized architecture (cost/latency).

🚀 Key Expertise & Added Value

1. High-Precision RAG Architectures

Implementation of cutting-edge pipelines: Hybrid Retrieval (BM25 + Vector Search), Reranking (BGE), and RRF (Reciprocal Rank Fusion).
Relevance optimization through Semantic Chunking and intent detection.
Management of complex and secure multi-tenant architectures.

2. Industrialization & Performance (Scale)

Code optimization for high availability: support for ~700 RPM.
Drastic reduction in latency and token costs (Advanced Prompt Engineering & Small Models).
Microservices architecture (Java/Spring Boot & Python/FastAPI) designed for scale.

3. AI Safety & Trust (Guardrails)

Real-time Guardrails system deployment: toxicity detection, irrelevance detection, language detection, and hallucination filtering.
Compliance with the requirements of highly regulated sectors (Banking/Insurance).

🏆 Why choose me?

"Real-World" Experience: I contributed to the first RAG-based virtual assistants in France (Helloïse / Telmi).

Dual Competence:AI Expertise (NLP, LLM) coupled with Software Engineering rigor (Hexagonal Architecture, Clean Code).

World Champion:Winner of the IWSLT 2022 challenge (Speech Recognition), proving deep mastery of SOTA models.

Let's move your AI project from a demonstrator to an industrial product.

Industry field of expertise

Languages

French
Fluent
English
Fluent
Arabic
Native or bilingual

Workplace preferences

Can work on-site

Paris (up to 20km)

bnpparibas
AI Engineer
BANKING AND INSURANCE
July 2025 - Today (11 months)
Montreuil, France
Contribution to the production deployment of two banking virtual assistants (HelloÏz, telmi) based on multi-tenant RAG architectures.

Deployment of AI systems supporting ~500 requests/minute for a user base of ~8 million users
Design of a hybrid RAG pipeline (BM25 + vector search + RRF + reranking) improving response relevance
Implementation of guardrail mechanisms via parallel calls to reduce latency
Performance optimization for banking production use
Development of data ingestion and structuring pipelines to ElasticSearch
Contribution to microservices architecture (APIs, deployment, monitoring)
Microservices Architecture Back-end Service Development Retrieval-Augmented Generation (RAG) Python Natural Language Processing (NLP)
bnpparibas
DATA SCIENTIST
BANKING AND INSURANCE
September 2023 - June 2025 (1 year and 9 months)
Montreuil, France
Exploration of AI solutions (NLP, LLM, Graphs) to automate legal compliance, CSR analysis, and permanent control, transforming complex data into operational production levers.

*Permanent Control:Automation of the scoring, reading, and information extraction process for documents stored in SharePoint. The solution integrates advanced NLP analysis, significantly improving document processing.

*Mesis:Design and development of a Spring Boot-based backend web application for measuring companies' environmental impact. The platform automates CSR (Corporate Social Responsibility) report generation and integrates advanced dynamic reporting features with JasperReports.

*Smart Review:Development of a knowledge graph-based system to identify inconsistencies or contradictions between different web pages. Using LLMs and Named Entity Recognition (NER) models, I extracted entities, relationships, and semantic patterns to build an automated content validation solution. By utilizing graph databases (Neo4j), I optimized the storage and querying of extracted relationships, thereby improving entity disambiguation and relationship extraction performance.

*Legal Design:Development of an automated legal document verification system based on NLP to ensure compliance with drafting standards. The system detects issues such as passive voice usage, negations, and similarities between sections using a combination of LLMs, embeddings, and linguistic heuristics. This solution helps legal teams improve document clarity, identify ambiguous phrasing, and ensure regulatory compliance.
Java Python Natural Language Processing (NLP) Hugging Face LLM
Carrefour
software engineer
E-COMMERCE
September 2021 - September 2023 (2 years)
Massy, France
Design and implementation of the BFF (Backend for Frontend) pattern to reduce business logic within the mobile application.
Decoupling the microservices architecture from backend services, thereby improving mobile application performance.
Implementation of the hexagonal architecture for microservices development.
Development of load tests and simulated backend services (mock backend).
Adoption of an API design-first approach based on an OpenAPI specification.
Technologies: Java, SpringBoot, Gatling, Scala, Reactive Programming, ELK, Grafana, Azure, GCP
Java Microservices Architecture Spring Boot Docker Kubernetes