Santander AI Lab - Frontier Agentic AI Engineer

Description

The role is with the AI Lab inside the chief data and AI division of a large bank, based in Madrid, Spain.

The lab takes ideas from early prototype to production. It builds working prototypes, tests them on real data and hands the strongest ones over to scale up, working with Anthropic, Sakana AI, AWS, ICMAT, CMU, INRIA and other partners.

The lab's published research on arXiv (arXiv:2602.14606, arXiv:2602.02170) underpins its work.

The bank is in the middle of a large technology-led transformation, and the division is building its AI and data team to serve customers across many countries.

The bank expects every professional to take a proactive, responsible approach to risk.

Responsibilities

• Build and deploy production agentic AI systems with multi-agent orchestration, memory, planning, tool use and error recovery.
• Develop and fine-tune small and medium language models for regulated banking tasks, including custom evaluations and domain benchmarks.
• Design and implement MCP servers, A2A protocols and federated API layers so AI agents can work across the group's infrastructure in several countries.
• Run two-week prototype sprints covering hypothesis, architecture, code, a working demo and a one-page summary.
• Work with researchers, data scientists and business stakeholders to turn complex technical ideas into practical value for the bank, reworking existing assets where needed.
• Track new research and tools such as Harness Engineering, Kiro, Windsurf and Devin, and adopt them in the lab's workflow when they prove useful.
• Write clean, tested, observable code that the AI Science team can take into production, and own your code from start to finish.

Requirements

• 4, 8 years of software or AI engineering experience, including at least 2 years building and running LLM-powered systems in production, not notebooks.
• Hands-on experience designing and deploying multi-agent systems with memory, stateful orchestration, tool use, multi-step planning and graceful failure recovery.
• Experience adapting language models for specific domains using SFT, LoRA or RLHF, including dataset curation and evaluation design.
• A record of delivering complete systems on your own under tight deadlines, from architecture to production-ready code, with a portfolio of working systems.
• Experience building and consuming REST APIs and connecting AI systems to enterprise data, cloud services and third-party platforms.
• A bachelor's or master's degree in computer science, software engineering, mathematics, physics or a closely related technical field.
• Professional working proficiency in English and Spanish. English is used for all technical documentation, papers and partner communication; Spanish for daily team and stakeholder work.
• Advanced Python with clean, tested, production-grade code.
• Experience with agentic frameworks such as LangGraph, AutoGen, CrewAI or Semantic Kernel, having built multi-agent workflows with them.
• Familiarity with LLM APIs and open models (Anthropic Claude, OpenAI, Llama, Mistral), and an understanding of cost, latency and quality trade-offs.
• Experience deploying and monitoring AI systems on AWS (Bedrock, Lambda, SageMaker, S3).
• Experience designing evaluations, including LLM-as-judge, RAGAS, Promptfoo, Langfuse or equivalent.
• API development with FastAPI or equivalent, building production-ready APIs around AI systems.
• Working knowledge of Docker, Git and CI/CD pipelines.
• Preferred: prior experience in banking, fintech or another regulated industry.
• Preferred: exposure to Harness Engineering methods such as spec-driven development and AI-assisted software creation at scale.
• Preferred: a master's or equivalent advanced qualification in AI, machine learning or a related field.
• Preferred: MCP server design and implementation.
• Preferred: SLM fine-tuning pipelines using vLLM, Ollama, Unsloth or equivalent.
• Preferred: knowledge graphs or GraphRAG for structured retrieval in complex domains.
• Preferred: A2A protocol, x402 or AP2 for agentic payments or agent-to-agent communication.
• Preferred: Kubernetes, Terraform or equivalent for production-scale deployment.

Conditions

Hybrid working, with some days remote and some days on site with your team, plus flexible hours.

Equal opportunities regardless of age, gender, disability, civil status, race, religion or sexual orientation.

Reasonable accommodations are available for people with disabilities.

Benefits

• A competitive salary with performance-based bonuses.
• Preferential banking terms, reduced loan interest rates and life insurance.
• Access to hundreds of courses and the bank's global learning programme, the company Open Academy.
• Opportunities to work across countries and follow varied career paths.
• Opportunities to publish research if the work merits it.
• A global wellness programme, BeHealthy.
• Childcare support and family programmes for each life stage.
• The company Contigo, offering legal, emotional and administrative advice for employees and their families.
• Gym or WellHub membership, medical centres at some sites, meal subsidy, parking, shuttle service from several points in Madrid, and employee discounts.