I started in full-stack work and kept following the harder problems — first down into the infrastructure that decides whether software ships and stays up, then into AI systems that have to behave in production, not just in a demo. Today I work across that whole line: RAG pipelines and LLM agents on one end, Kubernetes, OpenShift, and CI/CD on the other.
Most of what I've built is in real use — a mock-generation studio the Orange TV backend teams use at Sofrecom, a document assistant employees query at TEAMWILL, a CRM serving 2,000+ subscribed users, an ERP delivered across three OpenShift environments with GitOps. What I care about most is systems that are safe to hand to someone else: tested, documented, audited, and with a human in the loop where it matters.