> Markdown version of [/jobs/ext/4054776-staff-engineer-deployment-architecture-onprem](https://www.wearedevelopers.com/jobs/ext/4054776-staff-engineer-deployment-architecture-onprem). Every page supports `.md` or `Accept: text/markdown`. Links point to the HTML versions so they work for humans too. Agent guide: [/agents.md](https://www.wearedevelopers.com/agents.md). --- # Staff Engineer, Deployment Architecture onPrem - **Company:** Qevlar AI - **Location:** Paris, France (Remote available) - **Contract:** Permanent contract - **Skills:** Software Debugging, Public Key Infrastructure, Kubernetes - **Published:** October 20, 2026 - **Apply:** https://www.indeed.com/viewjob?jk=f1bc50d1a0f1aa2a ## About the Role * Deployed complex software into large enterprise datacentres, as a deployment architect, delivery lead, professional services engineer or technical program manager. You know what a six-week PKI ticket feels like. * A real technical floor. You read a values.yaml , debug a TLS handshake, explain what a proxy with a custom CA does to a container, and know what an OpenShift SCC is. You will not write the feature, but you will understand exactly why it broke. * Run several technical projects in parallel, each with its own stakeholders, timeline and definition of done, without losing any of them. * Serious writing skills. Documentation a stranger can execute without calling you is the bar we hold you to. * Composure in front of infrastructure architects, security officers and CISOs. These conversations are technical, adversarial by design, and decisive for the account. * Fluent English and French. ## Description * Own the pre-flight. Build and run the prerequisites assessment: Kubernetes distribution, GPU capacity, storage, identity provider, network egress, certificate authority. Every gap surfaces before install week, not during it. * Navigate the customer's organisation. Security reviews, firewall change requests, PKI, CAB cycles. You hold the schedule even though you control neither the resources nor the decisions. * Treat documentation as the product. We cannot log into our customers' environments. The install guide, the prerequisites matrix and the troubleshooting runbook are what these users actually consume. The acceptance test is a customer engineer completing an install without calling us. * Own the release train to customers. Version announcements, upgrade campaigns, migration planning across a fleet where every customer sits on a different release. * Run L3 triage. Qualify the issue, reproduce it from a support bundle, and escalate with everything engineering needs. You are the reason our engineers can debug an environment they will never see. * Close the loop. Every friction point you hit in the field becomes a documented requirement, a product ticket or a change to the bundle. The playbook gets shorter every quarter. * Cut the cycle time. Signature to production should shrink measurably across your first four deployments. That number is yours.