Agentic Platform Engineer

Job Description
We are Adevinta, a leading online classifieds group and champion for sustainable commerce across Europe and beyond. With digital marketplaces across many countries, our portfolio spans consumer goods, mobility, real estate, holiday rentals and jobs. Together, our marketplaces attract more than 120 million users every month and over a million businesses. Our locally loved marketplaces, including Marktplaats in the Netherlands, mobile.de in Germany, Kleinanzeigen in Germany and Spain, leboncoin in France and Subito in Italy, help everything and everyone find a new purpose by creating perfect matches – whether it’s a car, an apartment, a sofa or a new job. Our brands are supported by global Hubs in Barcelona and Amsterdam. Regarding this opportunity Anyone can wire an agent to a single repo and call it a demo. Wiring agents into the source control, CI/CD, secrets, observability, and approval flows of our marketplace engineering organisations — safely, reusably, and at controlled cost — is a different job. It's yours. Adevinta runs Leboncoin, Kleinanzeigen, Marktplaats, Subito, Mobile.de, and more. Real production systems, real users, real revenue. Not a lab. This new team will incubate our shared engine for making agentic engineering the default across marketplaces. Our engineers ship agent-built features inside marketplace teams. You build the platform underneath them — so what works in one pilot becomes a paved road for everyone else. If you've built internal developer platforms before, and you've been quietly waiting for the right time to apply that craft to agents, this is the role. Job Requirements Who you are Platform-engineering job first, with AI-runtime depth on top. We hire for shipped platforms used by our engineers, not for novelty.
• You have built internal developer platforms or developer tooling that real teams depend on. Not a side project — a platform with users, SLAs, and on-call rotations.
• You've designed and operated secure execution environments — sandboxing, runtime isolation, RBAC, secrets, audit logging — at production scale.
• You've shipped integrations across source control, CI/CD, ticketing, documentation, cloud, and observability systems. Polyglot estates don't scare you.
• You've built or operated tool gateways, adapters, or MCP-style integrations for agents — or the closest equivalent in the pre-agent era — and you understand the new failure modes.
• You instrument by default. Traces, telemetry, cost-per-workflow, eval pipelines — you don't ship a system you can't see into.
• You have an SRE mindset: reliability, rollback, rate limiting, graceful degradation, cost control are first-class concerns, not afterthoughts.
• You ship reusable templates and starter kits that teams actually adopt — not internal frameworks nobody uses.
• You've built or operated durable workflow systems. You understand persisted state, retries, idempotency, and recovery from partial failures, and can apply that experience to agent workflows that span tools, services, and human approvals.
• You're comfortable across modern infrastructure: containers, cloud (AWS or equivalent), IaC, CI/CD, APIs, scripting languages.
• You've supported many engineers or multiple teams as your customers — you know how to balance opinionated defaults with the inevitable special case. Job Responsibilities What you'll own The agent runtime Secure execution environments for agents. Sandboxing, runtime isolation, network policies, secrets management, RBAC, approvals. The boring-but-critical layer that makes "yes, you can let an agent touch our infra" a credible statement. The integration surface Adapters and gateways — MCP-style or otherwise — that let agents act on source control, CI/CD, ticketing, documentation, cloud, and observability systems across marketplaces. One opinionated way to do each integration, not ten flavours. The evaluation pipeline Golden tasks, graders, regression tests, release gates. Agentic workflows degrade silently without these. You build the plumbing that makes correctness measurable — and you make it cheap enough that teams actually use it. Observability and cost control Traces, telemetry, token usage, cost-per-workflow, rate limits, failure handling. Marketplace CTOs need to see what's happening, what it costs, and what to do when it breaks. You give them the dashboards and the controls. The durable execution layer Durable orchestration for longer-running agent workflows. Persisted workflow state, checkpointing, retries, timeouts, cancellation, and pause/resume for human approvals. You build the layer that lets workflows recover safely after failures and deployments, with recorded model and tool outcomes. Work can span minutes or days and still be recoverable. Reusable scaffolding Templates, starter kits, wrapper scripts, workflow automation. The patterns the Applied Agentic Engineers prove in marketplace pilots — you turn them into reusable platform components every team can adopt in a day. Job Benefits

What we offer
• A founding seat — Day-zero platform team. ExCo-backed mandate, real funding. You shape what gets built and what doesn't.
• Frontier latitude — Authority to pick model providers, integrations, and partners. We don't gate experimentation.
• Tight feedback loop — Our engineering teams are our closest internal customers. Real users from week one.
• Comp & flexibility — Compensation aligned with the seniority of the role and the talent density we're optimising for. Location and incentive structure to be shaped together.
• Working mode — Default in-person at our office in Barcelona. Platform work benefits from same-room debugging.

How to apply
Send a short note (no need for a cover letter) along your CV telling us:
• The internal platform you're proudest of. Who used it, what it did, and what you'd build differently now.
• Your view on what the right v0 of an agentic platform looks like inside a multi-stack, multi-cloud organisation.
• One thing in this job spec you'd push back on.