About Customs Support Group Customs Support Group (CSG) is the European market leader in customs services, operating across 15 countries.
CSG is owned by private equity and a fast-growing, dynamic, data-driven company, dedicated to expanding its presence through strong organic growth and strategic acquisitions.
Its key priorities include digital transformation, operational excellence and customer experience, all aimed at driving growth and enhancing efficiency.
You will design, test and implement practical and compliant agentic AI solutions on top of foundation models document understanding, classification support, retrieval over our own procedures and tariff material, and agents that carry multi-step operational work end to end.
Develop and customise AI solutions for the market, from prototype through to production ownership.
- Identify automation and decision-support opportunities across customs and business processes, and size them honestly before building.
- Monitor performance, inference cost, latency and business impact.
- Collaborate with local teams, Group functions and technology partners.
- Contribute to privacy, security and AI governance compliance, working with Group Legal and IT Security.
1-2 years AI engineering, software engineering or intelligent automation, including LLM or AI features that real users depended on.




- Strong Python and SQL, with production code ownership.
Demonstrable experience integrating large language models into production systems — prompt and context design, retrieval-augmented generation, structured output, and handling failure gracefully.
- Practical experience with at least one major cloud AI platform (Azure AI, AWS Bedrock, or Google Vertex).
- Solid full stack backend engineering fundamentals — APIs, data pipelines, testing, CI/CD.
Professional fluency in English.
Evaluation engineering: building eval harnesses and regression suites for non-deterministic systems, LLM-as-judge approaches, and human-in-the-loop review design.
Context engineering: designing and version-controlling what a model sees — retrieval strategy, memory, tool descriptions, and policy layers that constrain what an agent is allowed to do.
- Agent orchestration frameworks and standards (LangGraph, MCP, Temporal or equivalent), and experience with long-running or multi-step agent workflows.
- Observability and cost control for inference at volume.
Professional growth opportunities, including a budget for conferences, certification and model/compute costs for your own experiments.
- Permanent role.
Hybrid working.

📌 Customs Ai Integration (Zaragoza)
🏢 Customs Support
📍 Zaragoza