Scania is undergoing a transformation from being a leading supplier of trucks, buses, and engines to providing complete and sustainable transport solutions. Together with TRATON and our sister brands MAN, Volkswagen Truck & Bus, and International, we are working to shape the future of mobility with innovative and environmentally conscious solutions. Our values – customer first, respect, team spirit, responsibility, and elimination of waste – are at the heart of everything we do. Together, we are at the forefront of creating a sustainable future. Role Summary Embedded in Scania Logistics to build and scale governed agentic workflows across order-to-delivery. Autonomous logistics does not fail on ambition or ideas; it fails on the last mile between a proven demo and a governed production system. That last mile is exactly what this role owns, and each competency below is chosen against the goal, not against a generic job ladder. Design, build and operate production agentic workflows that take real order-to-delivery work, read approved context and enterprise data, use tools to act, and hand to humans at defined control points. The role turns proven proof lanes into reusable, governed capability that stays Scania-owned, and is measured on production outcomes rather than prototypes.Job Responsibilities Build agent workflows. Tool and function calling, orchestration and reasoning loops, state and memory, not just prompts. Engineer tools and integrations. Define tool and API contracts with permissions, retries and error handling, so agents act safely on order-to-delivery systems. Own evaluation. Build regression and trajectory evals tool choice, policy compliance, outcome quality and wire them into CI as quality gates. Guardrails and governance. Approvals for irreversible actions, audit logging, sensitive-data handling and prompt-injection defense, written as policy in code. Human-in-the-loop by design. Decide where humans review or approve, define escalation, and instrument the agent-to-human handoff governed autonomy. Ground the agents. Retrieval, memory and context engineering, including MCP, so agents are grounded in Scania data with citations. Productionize and operate. Observability and tracing, cost and latency control, safe rollout and incident response across the lane portfolio. Who You Are What matters most: the ability to build governed agentic workflows and to evaluate them honestly. Everything else strengthens the fit but is not required. Built with agents. Hands-on with LLM agents and tools; you know what holds up in production, not just demos. Evaluation as a discipline. You design evals and treat them as the hardest and most important part of the work. The modern agent stack. Comfortable with orchestration the OpenAI Agents and Anthropic SDKs, MCP, vector search and RAG, tracing and eval tooling. Good to have: a logistics or supply-chain operations background, and strong software engineering Python, APIs, distributed-systems thinki
