MACIF, Niort, France

• Designed, built and operate iard-run, an internal runbook web application (Python/Flask, Gunicorn) used by around twenty people in L2 support and P&C business run teams; AI-assisted development, deployed on MACIF infrastructure.
• Full production ownership of the application: architecture, application security and CVE management, CI/CD, deployment, monitoring and alerting (Dynatrace, Datadog) and continuous production operations.
• Designed and delivered PV Center, an internal Python/Flask (Gunicorn) application that centralises, keeps the history of and standardises acceptance-test sign-off reports before production releases, intended for about ten teams in the P&C scope (release managers, Scrum Masters, lead developers, business analysts, testers). Deployed on the enterprise OpenShift (Kubernetes) platform through the infra team's AutoDevOps pipeline, secrets from the internal configuration repository, Dynatrace monitoring, validated reports archived as PDF/A on S3.
• Proof of concept using LLMs to analyse bodily-injury claim documents and assess claim severity: multi-cloud pipeline from AWS S3 to Google Cloud Storage (Storage Transfer Service), Python processing on Cloud Run, Vertex AI models. Successful POC: feasibility demonstrated.
• Set up development environments on AWS EC2 (Dokemon, adapted in-house) used by at least twenty developers to pull and run the SaaS application code.
• Run around 30 to 40 Guidewire production releases and dry runs per year for the P&C scope.
• Maintain and evolve the P&C DevOps toolchain: Ansible/AWX automation, GitLab CI, secrets management with Bitwarden and OpenBao.
• Continuous production monitoring, on-call duty at least once a month, incident resolution with L2 teams and troubleshooting of non-production anomalies.

Conserto, on assignment at MACIF (GOING programme), Niort, France

• Set up the Guidewire SaaS environments and monitoring of the Guidewire Cloud Platform (GWCP).
• Improved DevOps tooling and automated operations tasks.
• Handled user incidents during non-regression testing (VNR) and dry-run phases.
• Wrote technical documentation.

Conserto, on assignment at MAIF (Digital Factory), Niort, France

• Maintained and evolved infrastructure and DevOps tooling; automated operations tasks.
• Coordinated deployments across all environments and maintained application configuration in non-production environments.
• Analysed and resolved production incidents; continuously improved monitoring.
• Administered the Jahia CMS.

Capgemini, on assignment at MACIF, Niort, France

• Scope: business intelligence, accounting, management control, HR and Big Data applications.
• Deployed and technically tested application components across all environments; day-to-day operations of business applications.
• Incident and request management; wrote operating procedures for L1/L2 support teams.
• Big Data specialisation: data lake ingestion acceptance testing, incident analysis, deployments.

Digicom 360° (founder), Le Monde Inuit

Audiovisual production and aerial filming, before moving into IT in late 2016.

TSK Arena, esports association

• AI-assisted development and operations of full-stack esports team management applications (scheduling, members, association tracking).
• Operate the development infrastructure for a team of 4 developers: Forgejo with CI/CD runners, Traefik, Uptime Kuma, Grafana/Loki/Promtail, Homarr dashboard.
• Proof-of-concept AI code-review agent: a Forgejo webhook triggers analysis by a local LLM (Ollama/LangChain), which posts a review comment or opens a merge request; every action requires human approval.
• Host and support the Discord bots built by the team.

Stack: Python, Go, PostgreSQL, Docker, Forgejo, Ollama, LangChain

Homelab and self-hosted platform

• Hybrid on-premise / Scaleway cloud infrastructure hosting 20+ services (personal cloud, monitoring stack, AI-assisted applications, TSK Arena infrastructure); this website and resume are hosted on it.
• Full GitOps workflow: Docker Compose, Traefik reverse proxy, Authentik SSO, Terraform and Ansible provisioning, Grafana monitoring, automated backups.

Stack: Debian, Docker, Traefik, Terraform, Ansible, Scaleway, Grafana, Forgejo, Authentik

• Design and deployment of web applications and internal tools (Python/Flask, FastAPI, Go, React/TypeScript, PostgreSQL) with an AI-augmented method: architecture designed before code, systematic review of generated code, built-in security (CSP, SQL injection prevention), logging and monitoring. Open-source example: JamGroove.

Stack: Python, Flask, FastAPI, Go, React, TypeScript, PostgreSQL, Claude Code