I build reliable RAG systems for Air Traffic Management.
Toulouse, France
About
I build AI systems for Air Traffic Management, a domain where reliability is non-negotiable.
I lead the development of an on-premise RAG platform that gives ATM teams fast, trustworthy access to thousands of pages of operational documentation. It is built as microservices (Python, FastAPI, Qdrant, PostgreSQL), with hybrid retrieval, continuous evaluation, and deployment on Docker and OpenShift.
Beyond the code, I act as technical lead: I run the team's technology watch, mentor apprentices, and have presented our work at Airspace World in Lisbon.
Focus
Production RAG and GenAI
Domain
Air Traffic Management
Company
Sopra Steria
Base
Toulouse, France
Experience
Sept 2024 – Present
AI & GenAI Engineer · Sopra Steria, Toulouse
Lead development of an on-premise RAG platform for Air Traffic Management, designed as microservices.
Built the evaluation pipeline that measures answer quality continuously (RAGAS, LangSmith).
Deploy and operate the platform with Docker and OpenShift.
Technical referent: technology watch, mentoring apprentices, presenting at Airspace World (Lisbon).
Python
FastAPI
Qdrant
PostgreSQL
RAGAS
LangSmith
Docker
OpenShift
Sept 2023 – Sept 2024
AI Engineer, R&D apprentice · Sopra Steria
Optimised YOLOv5 with pruning, quantization and distillation, with under 5% accuracy loss.
Exported models to ONNX, TensorRT, OpenVINO and TFLite for different deployment targets.
Measured the environmental impact of training and inference with CodeCarbon.
YOLOv5
ONNX
TensorRT
OpenVINO
TFLite
CodeCarbon
Jun – Sep 2023
Developer intern · Citec Ingénieurs Conseils
Automated ArcGIS Pro workflows and worked with GTFS transit data.
