Senior NLP Engineer — Public Service Automation

Nadia Kowalski

Full-time · Senior · Paris

This role is no longer accepting applications

Similar open roles you can apply to now:

About the role

We are a French public-sector technology agency operating under a mandate to modernise the digital infrastructure of government services. Our teams build and maintain the systems that process citizen requests, classify administrative documents, and route queries across seventeen ministerial departments. We operate under RGPD and French national data sovereignty requirements, which means that every system we build must be auditable, explainable, and run entirely within sovereign cloud infrastructure. We do not use third-party LLM APIs. We fine-tune and deploy our own models on our own infrastructure. The NLP engineering team is currently a group of nine engineers and three research scientists. We are growing by four positions this year as part of a multi-year investment in AI-assisted public administration. The role we are hiring for now focuses on document classification and information extraction: building systems that can reliably categorise citizen correspondence by type, extract structured data from unstructured administrative forms, and route documents to the correct department with an accuracy level that reduces manual handling costs significantly. The stakes are not abstract — our systems directly affect whether a citizen's request is processed in two days or six weeks.

Responsibilities

Design and train document classification models for citizen correspondence across ministerial domains
Build information extraction pipelines that identify and structure key fields from administrative forms with high reliability
Develop evaluation frameworks that measure accuracy at the level required for production deployment in public service workflows
Collaborate with legal and security teams to ensure all systems meet RGPD requirements and national data sovereignty standards
Contribute to the team's shared NLP infrastructure: training pipelines, model versioning, and serving standards

Requirements

7+ years of NLP engineering with production deployment experience on classification, extraction, or information retrieval tasks
Hugging Face — fine-tuning transformer models on domain-specific French and multilingual text corpora, evaluation pipelines, and Hub integration
PyTorch for custom training loops, model optimisation, and understanding what is happening beneath the library abstractions
Docker for reproducible training and serving environments that meet our internal security and compliance build standards
Kubernetes for deploying NLP inference services within our on-premises sovereign cloud infrastructure
SQL for working with our document metadata, routing logs, and quality validation datasets
French language proficiency at a professional level — the systems process French administrative text and the team works in French

Benefits

€72,000 – €90,000 gross annual salary — public sector pay scale with transparent progression bands
Full remote within France — team days in Paris quarterly
50 days RTT and annual leave combined — French public sector conditions
Pension scheme, complementary health insurance fully covered
Job security and mission that outlasts any startup runway