AI engineering · DevSecOps · Cloud architecture
Secure AI systems — from architecture to production.
I’m Paul-Émile Arnaly, an engineer based in Munich. I design and build AI-powered software and the DevSecOps platforms it runs on — for organisations where security and compliance are not optional.
01 What I do
AI
AI engineering
Agents, LLM integrations and MCP servers that connect AI to real systems — AI-assisted delivery with a human in the loop.
OPS
DevSecOps & cloud
CI/CD with GitHub Actions, Kubernetes and Terraform on AWS and Scaleway — platforms that are fast, reproducible and cost-aware.
SEC
Security & compliance
ISO 27001, DORA and audit evidence built into delivery from the start, not bolted on before the review.
02 Selected projects
Large insurance group
AI delivery orchestrator
Claude running in GitHub Actions turns a ticket into scoped sub-tasks and pull requests across the affected services. Every change passes a two-pass AI code review and a human approval before it ships.
Personal
Secure agent platform
One hub runs about ten MCP servers for all AI sessions over token-protected localhost connections. Secrets stay in the operating system’s credential store and never reach the model.
Large insurance group
CI/CD & cloud modernisation
Jenkins replaced by reusable GitHub Actions across about fifteen services; build-once, promote-by-registry releases to two AWS regions; leaner test stages; DORA and audit evidence.
AI MedCare start-up · CISO
Cloud & ISMS from scratch
Kubernetes on AWS and Scaleway as Terraform code, an ISO 27001-aligned information security management system and identity management in Microsoft Entra ID.
Side project
Browser Injector
A browser extension for Firefox, Chrome, Edge and Opera that injects custom CSS and JavaScript into web pages. Built with Angular.
03 Experience
École Centrale Paris & TU Munich — double degree · ISO 27001 · FR · DE · IT · EN
04 Contact
Happy to talk about AI, DevSecOps and secure cloud architecture.