Description

Basé à Paris, j’accompagne les entreprises dans leurs projets d’intelligence artificielle — de la recherche à la mise en production.

I’m an AI Research and MLOps Engineer with over 6 years of experience bridging research and production. I’ve designed and deployed domain-specific LLM, multimodal, and agentic AI systems — from fine-tuning models and building RAG pipelines to delivering scalable inference APIs and vector-based retrieval backends.

My background spans both applied AI and research — from multi-agent reinforcement learning at Orange Innovation to domain-specific LLM architectures at Lablabee. I focus on delivering reliable, high-impact AI systems that move smoothly from prototype to production.

Open for freelance missions involving LLMs, multimodal AI, RAG architectures, agentic systems, and MLOps pipelines.

Languages

English
Native or bilingual
French
Conversational

Workplace preferences

Can work on-site

Paris (up to 50km)

Lablabee
AI Lead – Research & LLM Engineering
November 2024 - October 2025 (11 months)
Paris, France
• Embedding Research (Telembed). Built a telecom-adaptive embedding framework with query–passage generation, FAISS-based hard negative mining, and two-stage fine-tuning (MNLR → Triplet-Loss), improving retrieval metrics (R@1, MRR) by 35% on 3GPP datasets.

• Agentic & RAG Systems. Designed autonomous multi-agent workflows for telecom RAG maintenance (importer, evaluator, curator agents), integrating FAISS, BGE embeddings, and LLM reasoning for automated evaluation and curation.

• LLM Inference & APIs. Implemented a low-latency streaming API (FastAPI, WebSocket/SSE) with a dynamic driver pool, cancellation support, and real-time metrics; deployed via CI/CD on self-hosted and AWS Bedrock environments.

• Data Pipelines & Retrieval. Developed end-to-end ingestion and retrieval pipelines (MongoDB Atlas, FAISS) with observability and CI/CD automation; enabled multi-source RAG retrieval for 3GPP and training documents.
Machine Learning Engineering FastAPI / Backend Development A/B Testing Large Language Models (LLMs) Retrieval-Augmented Generation (RAG)
Orange SA
AI Research Engineer
TECH
February 2023 - October 2024 (1 year and 8 months)
Paris, France
• MARL research: Developed a multi-agent RL framework with vertical (intra-domain) and horizontal (inter-domain) learning for network service placement; applied PPO with curriculum learning.

• LLMs & intent-based management: Built an intent-driven 5G/6G service management prototype with an NLP chatbot resolver; collaborated with Nokia Bell Labs on telecom-specific LLM use cases (NER, sentiment, negation), showing accuracy improvements.

• Forecasting: Developed state-of-the-art prediction models for upcoming service intents and resource demand, enabling proactive VNF placement and load balancing in simulated network environments.
Natural Language Processing (NLP) Large Language Models (LLMs) Machine Learning Engineering Reinforcement Learning
CNRS-
Marie-Curie PhD Fellow
RESEARCH
September 2019 - November 2022 (3 years and 2 months)
Paris, France
• Outage modeling: Proposed a stochastic geometry framework to analyze UAV communication outage probability; derived tractable expressions for optimal UAV altitude under LoS/NLoS conditions.

• Federated RL for localization: Designed a federated + reinforcement learning framework enabling multiple UAVs to localize ground users with faster convergence and reduced error collaboratively.

• Trajectory optimization: Developed deep reinforcement learning methods for UAV path planning to maximize communication throughput under
Reinforcement Learning Model Optimization Deep Learning Simulations